yyc12345/stb
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Merge pull request #7 from nothings/resample

Browse Source 
Resample
This commit is contained in:
Jorge Rodriguez 2014-09-13 12:31:18 -07:00
commit a0d4f79074
5 changed files with 1630 additions and 1387 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

446
stb_image_resize.h
View File

@ -82,7 +82,7 @@
and 99% opaque black produces 50% transparent dark green when and 99% opaque black produces 50% transparent dark green when
non-premultiplied, whereas premultiplied it produces 50% non-premultiplied, whereas premultiplied it produces 50%
transparent near-black. The former introduces green energy transparent near-black. The former introduces green energy
that doesn't exist in the source.) that doesn't exist in the source image.)
2. Artists should not edit premultiplied-alpha images; artists 2. Artists should not edit premultiplied-alpha images; artists
want non-premultiplied alpha images. Thus, art tools generally output want non-premultiplied alpha images. Thus, art tools generally output
@ -247,10 +247,10 @@ STBIRDEF int stbir_resize_uint8_srgb_edgemode(const unsigned char *input_pixels
typedef enum typedef enum
{ {
STBIR_FILTER_DEFAULT = 0, // use same filter type that easy-to-use API chooses STBIR_FILTER_DEFAULT = 0, // use same filter type that easy-to-use API chooses
STBIR_FILTER_BOX = 1, // Actually a trapezoid. See https://developer.nvidia.com/content/non-power-two-mipmapping STBIR_FILTER_BOX = 1, // A trapezoid w/1-pixel wide ramps, same result as box for integer scale ratios
STBIR_FILTER_TRIANGLE = 2, // On upsampling, produces same results as bilinear texture filtering STBIR_FILTER_TRIANGLE = 2, // On upsampling, produces same results as bilinear texture filtering
STBIR_FILTER_CUBIC = 3, // A cubic b-spline: why this one ???????????? STBIR_FILTER_CUBICBSPLINE = 3, // The cubic b-spline (aka Mitchell-Netrevalli with B=1,C=0), gaussian-esque
STBIR_FILTER_CATMULLROM = 4, // interpolating cubic spline STBIR_FILTER_CATMULLROM = 4, // An interpolating cubic spline
STBIR_FILTER_MITCHELL = 5, // Mitchell-Netrevalli filter with B=1/3, C=1/3 STBIR_FILTER_MITCHELL = 5, // Mitchell-Netrevalli filter with B=1/3, C=1/3
} stbir_filter; } stbir_filter;
@ -488,6 +488,16 @@ typedef struct
float* horizontal_buffer; float* horizontal_buffer;
// cache these because ceil/floor are inexplicably showing up in profile
int horizontal_coefficient_width;
int vertical_coefficient_width;
int horizontal_filter_pixel_width;
int vertical_filter_pixel_width;
int horizontal_filter_pixel_margin;
int vertical_filter_pixel_margin;
int horizontal_num_contributors;
int vertical_num_contributors;
int ring_buffer_length_bytes; // The length of an individual entry in the ring buffer. The total number of ring buffers is stbir__get_filter_pixel_width(filter) int ring_buffer_length_bytes; // The length of an individual entry in the ring buffer. The total number of ring buffers is stbir__get_filter_pixel_width(filter)
int ring_buffer_first_scanline; int ring_buffer_first_scanline;
int ring_buffer_last_scanline; int ring_buffer_last_scanline;
@ -555,6 +565,10 @@ static float stbir__srgb_uchar_to_linear_float[256] = {
}; };
// sRGB transition values, scaled by 1<<28 // sRGB transition values, scaled by 1<<28
// note that if you only scaled by 1<<16, all the values would be 4K smaller,
// so [1] would be ~10, and so that would have around 5% error (10 +- 0.5)
// at the boundary between uint8 0 and 1. This also means that a 64K-entry table
// would have the same 5% error there.
static int stbir__srgb_offset_to_linear_scaled[256] = static int stbir__srgb_offset_to_linear_scaled[256] =
{ {
0, 40738, 122216, 203693, 285170, 366648, 448125, 529603, 0, 40738, 122216, 203693, 285170, 366648, 448125, 529603,
@ -749,7 +763,7 @@ stbir__inline static int stbir__use_height_upsampling(stbir__info* stbir_info)
// This is the maximum number of input samples that can affect an output sample // This is the maximum number of input samples that can affect an output sample
// with the given filter // with the given filter
stbir__inline static int stbir__get_filter_pixel_width(stbir_filter filter, float scale) static int stbir__get_filter_pixel_width(stbir_filter filter, float scale)
{ {
STBIR_ASSERT(filter != 0); STBIR_ASSERT(filter != 0);
STBIR_ASSERT(filter < STBIR__ARRAY_SIZE(stbir__filter_info_table)); STBIR_ASSERT(filter < STBIR__ARRAY_SIZE(stbir__filter_info_table));
@ -760,34 +774,14 @@ stbir__inline static int stbir__get_filter_pixel_width(stbir_filter filter, floa
return (int)ceil(stbir__filter_info_table[filter].support(scale) * 2 / scale); return (int)ceil(stbir__filter_info_table[filter].support(scale) * 2 / scale);
} }
stbir__inline static int stbir__get_filter_pixel_width_horizontal(stbir__info* stbir_info)
{
return stbir__get_filter_pixel_width(stbir_info->horizontal_filter, stbir_info->horizontal_scale);
}
stbir__inline static int stbir__get_filter_pixel_width_vertical(stbir__info* stbir_info)
{
return stbir__get_filter_pixel_width(stbir_info->vertical_filter, stbir_info->vertical_scale);
}
// This is how much to expand buffers to account for filters seeking outside // This is how much to expand buffers to account for filters seeking outside
// the image boundaries. // the image boundaries.
stbir__inline static int stbir__get_filter_pixel_margin(stbir_filter filter, float scale) static int stbir__get_filter_pixel_margin(stbir_filter filter, float scale)
{ {
return stbir__get_filter_pixel_width(filter, scale) / 2; return stbir__get_filter_pixel_width(filter, scale) / 2;
} }
stbir__inline static int stbir__get_filter_pixel_margin_horizontal(stbir__info* stbir_info) static int stbir__get_coefficient_width(stbir_filter filter, float scale)
{
return stbir__get_filter_pixel_width(stbir_info->horizontal_filter, stbir_info->horizontal_scale) / 2;
}
stbir__inline static int stbir__get_filter_pixel_margin_vertical(stbir__info* stbir_info)
{
return stbir__get_filter_pixel_width(stbir_info->vertical_filter, stbir_info->vertical_scale) / 2;
}
stbir__inline static int stbir__get_coefficient_width(stbir_filter filter, float scale)
{ {
if (stbir__use_upsampling(scale)) if (stbir__use_upsampling(scale))
return (int)ceil(stbir__filter_info_table[filter].support(1 / scale) * 2); return (int)ceil(stbir__filter_info_table[filter].support(1 / scale) * 2);
@ -795,7 +789,7 @@ stbir__inline static int stbir__get_coefficient_width(stbir_filter filter, float
return (int)ceil(stbir__filter_info_table[filter].support(scale) * 2); return (int)ceil(stbir__filter_info_table[filter].support(scale) * 2);
} }
stbir__inline static int stbir__get_contributors(float scale, stbir_filter filter, int input_size, int output_size) static int stbir__get_contributors(float scale, stbir_filter filter, int input_size, int output_size)
{ {
if (stbir__use_upsampling(scale)) if (stbir__use_upsampling(scale))
return output_size; return output_size;
@ -803,67 +797,31 @@ stbir__inline static int stbir__get_contributors(float scale, stbir_filter filte
return (input_size + stbir__get_filter_pixel_margin(filter, scale) * 2); return (input_size + stbir__get_filter_pixel_margin(filter, scale) * 2);
} }
stbir__inline static int stbir__get_horizontal_contributors(stbir__info* info) static int stbir__get_total_horizontal_coefficients(stbir__info* info)
{ {
return stbir__get_contributors(info->horizontal_scale, info->horizontal_filter, info->input_w, info->output_w); return info->horizontal_num_contributors
}
stbir__inline static int stbir__get_vertical_contributors(stbir__info* info)
{
return stbir__get_contributors(info->vertical_scale, info->vertical_filter, info->input_h, info->output_h);
}
stbir__inline static int stbir__get_total_horizontal_coefficients(stbir__info* info)
{
return stbir__get_horizontal_contributors(info)
* stbir__get_coefficient_width (info->horizontal_filter, info->horizontal_scale); * stbir__get_coefficient_width (info->horizontal_filter, info->horizontal_scale);
} }
stbir__inline static int stbir__get_total_vertical_coefficients(stbir__info* info) static int stbir__get_total_vertical_coefficients(stbir__info* info)
{ {
return stbir__get_vertical_contributors(info) return info->vertical_num_contributors
* stbir__get_coefficient_width (info->vertical_filter, info->vertical_scale); * stbir__get_coefficient_width (info->vertical_filter, info->vertical_scale);
} }
stbir__inline static stbir__contributors* stbir__get_contributor(stbir__contributors* contributors, int n) static stbir__contributors* stbir__get_contributor(stbir__contributors* contributors, int n)
{ {
return &contributors[n]; return &contributors[n];
} }
stbir__inline static stbir__contributors* stbir__get_horizontal_contributor(stbir__info* stbir_info, int n)
{
STBIR__DEBUG_ASSERT(n >= 0 && n < stbir__get_horizontal_contributors(stbir_info));
return stbir__get_contributor(stbir_info->horizontal_contributors, n);
}
stbir__inline static stbir__contributors* stbir__get_vertical_contributor(stbir__info* stbir_info, int n)
{
STBIR__DEBUG_ASSERT(n >= 0 && n < stbir__get_vertical_contributors(stbir_info));
return stbir__get_contributor(stbir_info->vertical_contributors, n);
}
// For perf reasons this code is duplicated in stbir__resample_horizontal_upsample/downsample, // For perf reasons this code is duplicated in stbir__resample_horizontal_upsample/downsample,
// if you change it here change it there too. // if you change it here change it there too.
stbir__inline static float* stbir__get_coefficient(float* coefficients, stbir_filter filter, float scale, int n, int c) static float* stbir__get_coefficient(float* coefficients, stbir_filter filter, float scale, int n, int c)
{ {
int width = stbir__get_coefficient_width(filter, scale); int width = stbir__get_coefficient_width(filter, scale);
return &coefficients[width*n + c]; return &coefficients[width*n + c];
} }
stbir__inline static float* stbir__get_horizontal_coefficient(stbir__info* stbir_info, int n, int c)
{
STBIR__DEBUG_ASSERT(c >= 0 && c < stbir__get_coefficient_width(stbir_info->horizontal_filter, stbir_info->horizontal_scale));
STBIR__DEBUG_ASSERT(n >= 0 && n < stbir__get_total_horizontal_coefficients(stbir_info));
return stbir__get_coefficient(stbir_info->horizontal_coefficients, stbir_info->horizontal_filter, stbir_info->horizontal_scale, n, c);
}
stbir__inline static float* stbir__get_vertical_coefficient(stbir__info* stbir_info, int n, int c)
{
STBIR__DEBUG_ASSERT(c >= 0 && c < stbir__get_coefficient_width(stbir_info->vertical_filter, stbir_info->vertical_scale));
STBIR__DEBUG_ASSERT(n >= 0 && n < stbir__get_total_vertical_coefficients(stbir_info));
return stbir__get_coefficient(stbir_info->vertical_coefficients, stbir_info->vertical_filter, stbir_info->vertical_scale, n, c);
}
static int stbir__edge_wrap_slow(stbir_edge edge, int n, int max) static int stbir__edge_wrap_slow(stbir_edge edge, int n, int max)
{ {
switch (edge) switch (edge)
@ -1081,7 +1039,7 @@ static void stbir__normalize_downsample_coefficients(stbir__info* stbir_info, st
// Do this after normalizing because normalization depends on the n0/n1 values. // Do this after normalizing because normalization depends on the n0/n1 values.
for (j = 0; j < num_contributors; j++) for (j = 0; j < num_contributors; j++)
{ {
int range, max; int range, max, width;
skip = 0; skip = 0;
while (*stbir__get_coefficient(coefficients, filter, scale_ratio, j, skip) == 0) while (*stbir__get_coefficient(coefficients, filter, scale_ratio, j, skip) == 0)
@ -1098,9 +1056,10 @@ static void stbir__normalize_downsample_coefficients(stbir__info* stbir_info, st
range = contributors[j].n1 - contributors[j].n0 + 1; range = contributors[j].n1 - contributors[j].n0 + 1;
max = stbir__min(num_coefficients, range); max = stbir__min(num_coefficients, range);
width = stbir__get_coefficient_width(filter, scale_ratio);
for (i = 0; i < max; i++) for (i = 0; i < max; i++)
{ {
if (i + skip >= stbir__get_coefficient_width(filter, scale_ratio)) if (i + skip >= width)
break; break;
*stbir__get_coefficient(coefficients, filter, scale_ratio, j, i) = *stbir__get_coefficient(coefficients, filter, scale_ratio, j, i + skip); *stbir__get_coefficient(coefficients, filter, scale_ratio, j, i) = *stbir__get_coefficient(coefficients, filter, scale_ratio, j, i + skip);
@ -1160,7 +1119,7 @@ static float* stbir__get_decode_buffer(stbir__info* stbir_info)
{ {
// The 0 index of the decode buffer starts after the margin. This makes // The 0 index of the decode buffer starts after the margin. This makes
// it okay to use negative indexes on the decode buffer. // it okay to use negative indexes on the decode buffer.
return &stbir_info->decode_buffer[stbir__get_filter_pixel_margin_horizontal(stbir_info) * stbir_info->channels]; return &stbir_info->decode_buffer[stbir_info->horizontal_filter_pixel_margin * stbir_info->channels];
} }
#define STBIR__DECODE(type, colorspace) ((type) * (STBIR_MAX_COLORSPACES) + (colorspace)) #define STBIR__DECODE(type, colorspace) ((type) * (STBIR_MAX_COLORSPACES) + (colorspace))
@ -1179,10 +1138,10 @@ static void stbir__decode_scanline(stbir__info* stbir_info, int n)
stbir_edge edge_vertical = stbir_info->edge_vertical; stbir_edge edge_vertical = stbir_info->edge_vertical;
int in_buffer_row_offset = stbir__edge_wrap(edge_vertical, n, stbir_info->input_h) * input_stride_bytes; int in_buffer_row_offset = stbir__edge_wrap(edge_vertical, n, stbir_info->input_h) * input_stride_bytes;
const void* input_data = (char *) stbir_info->input_data + in_buffer_row_offset; const void* input_data = (char *) stbir_info->input_data + in_buffer_row_offset;
int max_x = input_w + stbir__get_filter_pixel_margin_horizontal(stbir_info); int max_x = input_w + stbir_info->horizontal_filter_pixel_margin;
int decode = STBIR__DECODE(type, colorspace); int decode = STBIR__DECODE(type, colorspace);
int x = -stbir__get_filter_pixel_margin_horizontal(stbir_info); int x = -stbir_info->horizontal_filter_pixel_margin;
// special handling for STBIR_EDGE_ZERO because it needs to return an item that doesn't appear in the input, // special handling for STBIR_EDGE_ZERO because it needs to return an item that doesn't appear in the input,
// and we want to avoid paying overhead on every pixel if not STBIR_EDGE_ZERO // and we want to avoid paying overhead on every pixel if not STBIR_EDGE_ZERO
@ -1296,7 +1255,7 @@ static void stbir__decode_scanline(stbir__info* stbir_info, int n)
if (!(stbir_info->flags & STBIR_FLAG_ALPHA_PREMULTIPLIED)) if (!(stbir_info->flags & STBIR_FLAG_ALPHA_PREMULTIPLIED))
{ {
for (x = -stbir__get_filter_pixel_margin_horizontal(stbir_info); x < max_x; x++) for (x = -stbir_info->horizontal_filter_pixel_margin; x < max_x; x++)
{ {
int decode_pixel_index = x * channels; int decode_pixel_index = x * channels;
@ -1320,7 +1279,7 @@ static void stbir__decode_scanline(stbir__info* stbir_info, int n)
if (edge_horizontal == STBIR_EDGE_ZERO) if (edge_horizontal == STBIR_EDGE_ZERO)
{ {
for (x = -stbir__get_filter_pixel_margin_horizontal(stbir_info); x < 0; x++) for (x = -stbir_info->horizontal_filter_pixel_margin; x < 0; x++)
{ {
for (c = 0; c < channels; c++) for (c = 0; c < channels; c++)
decode_buffer[x*channels + c] = 0; decode_buffer[x*channels + c] = 0;
@ -1350,7 +1309,7 @@ static float* stbir__add_empty_ring_buffer_entry(stbir__info* stbir_info, int n)
} }
else else
{ {
ring_buffer_index = (stbir_info->ring_buffer_begin_index + (stbir_info->ring_buffer_last_scanline - stbir_info->ring_buffer_first_scanline) + 1) % stbir__get_filter_pixel_width_vertical(stbir_info); ring_buffer_index = (stbir_info->ring_buffer_begin_index + (stbir_info->ring_buffer_last_scanline - stbir_info->ring_buffer_first_scanline) + 1) % stbir_info->vertical_filter_pixel_width;
STBIR__DEBUG_ASSERT(ring_buffer_index != stbir_info->ring_buffer_begin_index); STBIR__DEBUG_ASSERT(ring_buffer_index != stbir_info->ring_buffer_begin_index);
} }
@ -1367,12 +1326,12 @@ static void stbir__resample_horizontal_upsample(stbir__info* stbir_info, int n,
{ {
int x, k; int x, k;
int output_w = stbir_info->output_w; int output_w = stbir_info->output_w;
int kernel_pixel_width = stbir__get_filter_pixel_width_horizontal(stbir_info); int kernel_pixel_width = stbir_info->horizontal_filter_pixel_width;
int channels = stbir_info->channels; int channels = stbir_info->channels;
float* decode_buffer = stbir__get_decode_buffer(stbir_info); float* decode_buffer = stbir__get_decode_buffer(stbir_info);
stbir__contributors* horizontal_contributors = stbir_info->horizontal_contributors; stbir__contributors* horizontal_contributors = stbir_info->horizontal_contributors;
float* horizontal_coefficients = stbir_info->horizontal_coefficients; float* horizontal_coefficients = stbir_info->horizontal_coefficients;
int coefficient_width = stbir__get_coefficient_width(stbir_info->horizontal_filter, stbir_info->horizontal_scale); int coefficient_width = stbir_info->horizontal_coefficient_width;
for (x = 0; x < output_w; x++) for (x = 0; x < output_w; x++)
{ {
@ -1384,22 +1343,66 @@ static void stbir__resample_horizontal_upsample(stbir__info* stbir_info, int n,
int coefficient_counter = 0; int coefficient_counter = 0;
STBIR__DEBUG_ASSERT(n1 >= n0); STBIR__DEBUG_ASSERT(n1 >= n0);
STBIR__DEBUG_ASSERT(n0 >= -stbir__get_filter_pixel_margin_horizontal(stbir_info)); STBIR__DEBUG_ASSERT(n0 >= -stbir_info->horizontal_filter_pixel_margin);
STBIR__DEBUG_ASSERT(n1 >= -stbir__get_filter_pixel_margin_horizontal(stbir_info)); STBIR__DEBUG_ASSERT(n1 >= -stbir_info->horizontal_filter_pixel_margin);
STBIR__DEBUG_ASSERT(n0 < stbir_info->input_w + stbir__get_filter_pixel_margin_horizontal(stbir_info)); STBIR__DEBUG_ASSERT(n0 < stbir_info->input_w + stbir_info->horizontal_filter_pixel_margin);
STBIR__DEBUG_ASSERT(n1 < stbir_info->input_w + stbir__get_filter_pixel_margin_horizontal(stbir_info)); STBIR__DEBUG_ASSERT(n1 < stbir_info->input_w + stbir_info->horizontal_filter_pixel_margin);
switch (channels) {
case 1:
for (k = n0; k <= n1; k++)
{
int in_pixel_index = k * 1;
float coefficient = horizontal_coefficients[coefficient_group + coefficient_counter++];
STBIR__DEBUG_ASSERT(coefficient != 0);
output_buffer[out_pixel_index + 0] += decode_buffer[in_pixel_index + 0] * coefficient;
}
break;
case 2:
for (k = n0; k <= n1; k++)
{
int in_pixel_index = k * 2;
float coefficient = horizontal_coefficients[coefficient_group + coefficient_counter++];
STBIR__DEBUG_ASSERT(coefficient != 0);
output_buffer[out_pixel_index + 0] += decode_buffer[in_pixel_index + 0] * coefficient;
output_buffer[out_pixel_index + 1] += decode_buffer[in_pixel_index + 1] * coefficient;
}
break;
case 3:
for (k = n0; k <= n1; k++)
{
int in_pixel_index = k * 3;
float coefficient = horizontal_coefficients[coefficient_group + coefficient_counter++];
STBIR__DEBUG_ASSERT(coefficient != 0);
output_buffer[out_pixel_index + 0] += decode_buffer[in_pixel_index + 0] * coefficient;
output_buffer[out_pixel_index + 1] += decode_buffer[in_pixel_index + 1] * coefficient;
output_buffer[out_pixel_index + 2] += decode_buffer[in_pixel_index + 2] * coefficient;
}
break;
case 4:
for (k = n0; k <= n1; k++)
{
int in_pixel_index = k * 4;
float coefficient = horizontal_coefficients[coefficient_group + coefficient_counter++];
STBIR__DEBUG_ASSERT(coefficient != 0);
output_buffer[out_pixel_index + 0] += decode_buffer[in_pixel_index + 0] * coefficient;
output_buffer[out_pixel_index + 1] += decode_buffer[in_pixel_index + 1] * coefficient;
output_buffer[out_pixel_index + 2] += decode_buffer[in_pixel_index + 2] * coefficient;
output_buffer[out_pixel_index + 3] += decode_buffer[in_pixel_index + 3] * coefficient;
}
break;
default:
for (k = n0; k <= n1; k++) for (k = n0; k <= n1; k++)
{ {
int in_pixel_index = k * channels; int in_pixel_index = k * channels;
float coefficient = horizontal_coefficients[coefficient_group + coefficient_counter++]; float coefficient = horizontal_coefficients[coefficient_group + coefficient_counter++];
STBIR__DEBUG_ASSERT(coefficient != 0);
int c; int c;
STBIR__DEBUG_ASSERT(coefficient != 0);
for (c = 0; c < channels; c++) for (c = 0; c < channels; c++)
output_buffer[out_pixel_index + c] += decode_buffer[in_pixel_index + c] * coefficient; output_buffer[out_pixel_index + c] += decode_buffer[in_pixel_index + c] * coefficient;
} }
break;
}
} }
} }
@ -1408,17 +1411,109 @@ static void stbir__resample_horizontal_downsample(stbir__info* stbir_info, int n
int x, k; int x, k;
int input_w = stbir_info->input_w; int input_w = stbir_info->input_w;
int output_w = stbir_info->output_w; int output_w = stbir_info->output_w;
int kernel_pixel_width = stbir__get_filter_pixel_width_horizontal(stbir_info); int kernel_pixel_width = stbir_info->horizontal_filter_pixel_width;
int channels = stbir_info->channels; int channels = stbir_info->channels;
float* decode_buffer = stbir__get_decode_buffer(stbir_info); float* decode_buffer = stbir__get_decode_buffer(stbir_info);
stbir__contributors* horizontal_contributors = stbir_info->horizontal_contributors; stbir__contributors* horizontal_contributors = stbir_info->horizontal_contributors;
float* horizontal_coefficients = stbir_info->horizontal_coefficients; float* horizontal_coefficients = stbir_info->horizontal_coefficients;
int coefficient_width = stbir__get_coefficient_width(stbir_info->horizontal_filter, stbir_info->horizontal_scale); int coefficient_width = stbir_info->horizontal_coefficient_width;
int filter_pixel_margin = stbir__get_filter_pixel_margin_horizontal(stbir_info); int filter_pixel_margin = stbir_info->horizontal_filter_pixel_margin;
int max_x = input_w + filter_pixel_margin * 2; int max_x = input_w + filter_pixel_margin * 2;
STBIR__DEBUG_ASSERT(!stbir__use_width_upsampling(stbir_info)); STBIR__DEBUG_ASSERT(!stbir__use_width_upsampling(stbir_info));
switch (channels) {
case 1:
for (x = 0; x < max_x; x++)
{
int n0 = horizontal_contributors[x].n0;
int n1 = horizontal_contributors[x].n1;
int in_x = x - filter_pixel_margin;
int in_pixel_index = in_x * 1;
int max_n = n1;
int coefficient_group = coefficient_width * x;
for (k = n0; k <= max_n; k++)
{
int out_pixel_index = k * 1;
float coefficient = horizontal_coefficients[coefficient_group + k - n0];
STBIR__DEBUG_ASSERT(coefficient != 0);
output_buffer[out_pixel_index + 0] += decode_buffer[in_pixel_index + 0] * coefficient;
}
}
break;
case 2:
for (x = 0; x < max_x; x++)
{
int n0 = horizontal_contributors[x].n0;
int n1 = horizontal_contributors[x].n1;
int in_x = x - filter_pixel_margin;
int in_pixel_index = in_x * 2;
int max_n = n1;
int coefficient_group = coefficient_width * x;
for (k = n0; k <= max_n; k++)
{
int out_pixel_index = k * 2;
float coefficient = horizontal_coefficients[coefficient_group + k - n0];
STBIR__DEBUG_ASSERT(coefficient != 0);
output_buffer[out_pixel_index + 0] += decode_buffer[in_pixel_index + 0] * coefficient;
output_buffer[out_pixel_index + 1] += decode_buffer[in_pixel_index + 1] * coefficient;
}
}
break;
case 3:
for (x = 0; x < max_x; x++)
{
int n0 = horizontal_contributors[x].n0;
int n1 = horizontal_contributors[x].n1;
int in_x = x - filter_pixel_margin;
int in_pixel_index = in_x * 3;
int max_n = n1;
int coefficient_group = coefficient_width * x;
for (k = n0; k <= max_n; k++)
{
int out_pixel_index = k * 3;
float coefficient = horizontal_coefficients[coefficient_group + k - n0];
STBIR__DEBUG_ASSERT(coefficient != 0);
output_buffer[out_pixel_index + 0] += decode_buffer[in_pixel_index + 0] * coefficient;
output_buffer[out_pixel_index + 1] += decode_buffer[in_pixel_index + 1] * coefficient;
output_buffer[out_pixel_index + 2] += decode_buffer[in_pixel_index + 2] * coefficient;
}
}
break;
case 4:
for (x = 0; x < max_x; x++)
{
int n0 = horizontal_contributors[x].n0;
int n1 = horizontal_contributors[x].n1;
int in_x = x - filter_pixel_margin;
int in_pixel_index = in_x * 4;
int max_n = n1;
int coefficient_group = coefficient_width * x;
for (k = n0; k <= max_n; k++)
{
int out_pixel_index = k * 4;
float coefficient = horizontal_coefficients[coefficient_group + k - n0];
STBIR__DEBUG_ASSERT(coefficient != 0);
output_buffer[out_pixel_index + 0] += decode_buffer[in_pixel_index + 0] * coefficient;
output_buffer[out_pixel_index + 1] += decode_buffer[in_pixel_index + 1] * coefficient;
output_buffer[out_pixel_index + 2] += decode_buffer[in_pixel_index + 2] * coefficient;
output_buffer[out_pixel_index + 3] += decode_buffer[in_pixel_index + 3] * coefficient;
}
}
break;
default:
for (x = 0; x < max_x; x++) for (x = 0; x < max_x; x++)
{ {
int n0 = horizontal_contributors[x].n0; int n0 = horizontal_contributors[x].n0;
@ -1431,16 +1526,16 @@ static void stbir__resample_horizontal_downsample(stbir__info* stbir_info, int n
for (k = n0; k <= max_n; k++) for (k = n0; k <= max_n; k++)
{ {
int c;
int out_pixel_index = k * channels; int out_pixel_index = k * channels;
float coefficient = horizontal_coefficients[coefficient_group + k - n0]; float coefficient = horizontal_coefficients[coefficient_group + k - n0];
STBIR__DEBUG_ASSERT(coefficient != 0); STBIR__DEBUG_ASSERT(coefficient != 0);
int c;
for (c = 0; c < channels; c++) for (c = 0; c < channels; c++)
output_buffer[out_pixel_index + c] += decode_buffer[in_pixel_index + c] * coefficient; output_buffer[out_pixel_index + c] += decode_buffer[in_pixel_index + c] * coefficient;
} }
} }
break;
}
} }
static void stbir__decode_and_resample_upsample(stbir__info* stbir_info, int n) static void stbir__decode_and_resample_upsample(stbir__info* stbir_info, int n)
@ -1504,6 +1599,8 @@ static void stbir__encode_scanline(stbir__info* stbir_info, int num_pixels, void
} }
} }
#define STBIR__ROUND_INT(f) ((int) ((f)+0.5)) //#define STBIR__ROUND_INT(f) (floor((f)+0.5))
switch (decode) switch (decode)
{ {
case STBIR__DECODE(STBIR_TYPE_UINT8, STBIR_COLORSPACE_LINEAR): case STBIR__DECODE(STBIR_TYPE_UINT8, STBIR_COLORSPACE_LINEAR):
@ -1514,7 +1611,7 @@ static void stbir__encode_scanline(stbir__info* stbir_info, int num_pixels, void
for (n = 0; n < channels; n++) for (n = 0; n < channels; n++)
{ {
int index = pixel_index + n; int index = pixel_index + n;
((unsigned char*)output_buffer)[index] = (unsigned char)(floor(stbir__saturate(encode_buffer[index]) * 255 + 0.5f)); ((unsigned char*)output_buffer)[index] = (unsigned char) STBIR__ROUND_INT(stbir__saturate(encode_buffer[index]) * 255);
} }
} }
break; break;
@ -1531,7 +1628,7 @@ static void stbir__encode_scanline(stbir__info* stbir_info, int num_pixels, void
} }
if (!(stbir_info->flags&STBIR_FLAG_ALPHA_USES_COLORSPACE)) if (!(stbir_info->flags&STBIR_FLAG_ALPHA_USES_COLORSPACE))
((unsigned char*)output_buffer)[pixel_index + alpha_channel] = (unsigned char)(floor(stbir__saturate(encode_buffer[pixel_index + alpha_channel]) * 255+0.5f)); ((unsigned char*)output_buffer)[pixel_index + alpha_channel] = (unsigned char) STBIR__ROUND_INT(stbir__saturate(encode_buffer[pixel_index + alpha_channel]) * 255);
} }
break; break;
@ -1543,7 +1640,7 @@ static void stbir__encode_scanline(stbir__info* stbir_info, int num_pixels, void
for (n = 0; n < channels; n++) for (n = 0; n < channels; n++)
{ {
int index = pixel_index + n; int index = pixel_index + n;
((unsigned short*)output_buffer)[index] = (unsigned short)(floor(stbir__saturate(encode_buffer[index]) * 65535+0.5f)); ((unsigned short*)output_buffer)[index] = (unsigned short)STBIR__ROUND_INT(stbir__saturate(encode_buffer[index]) * 65535);
} }
} }
break; break;
@ -1556,11 +1653,11 @@ static void stbir__encode_scanline(stbir__info* stbir_info, int num_pixels, void
for (n = 0; n < channels; n++) for (n = 0; n < channels; n++)
{ {
int index = pixel_index + n; int index = pixel_index + n;
((unsigned short*)output_buffer)[index] = (unsigned short)(floor(stbir__linear_to_srgb(stbir__saturate(encode_buffer[index])) * 65535 + 0.5f)); ((unsigned short*)output_buffer)[index] = (unsigned short)STBIR__ROUND_INT(stbir__linear_to_srgb(stbir__saturate(encode_buffer[index])) * 65535);
} }
if (!(stbir_info->flags&STBIR_FLAG_ALPHA_USES_COLORSPACE)) if (!(stbir_info->flags&STBIR_FLAG_ALPHA_USES_COLORSPACE))
((unsigned short*)output_buffer)[pixel_index + alpha_channel] = (unsigned short)(floor(stbir__saturate(encode_buffer[pixel_index + alpha_channel]) * 65535 + 0.5f)); ((unsigned short*)output_buffer)[pixel_index + alpha_channel] = (unsigned short)STBIR__ROUND_INT(stbir__saturate(encode_buffer[pixel_index + alpha_channel]) * 65535);
} }
break; break;
@ -1573,7 +1670,7 @@ static void stbir__encode_scanline(stbir__info* stbir_info, int num_pixels, void
for (n = 0; n < channels; n++) for (n = 0; n < channels; n++)
{ {
int index = pixel_index + n; int index = pixel_index + n;
((unsigned int*)output_buffer)[index] = (unsigned int)(floor(((double)stbir__saturate(encode_buffer[index])) * 4294967295 + 0.5f)); ((unsigned int*)output_buffer)[index] = (unsigned int)STBIR__ROUND_INT(((double)stbir__saturate(encode_buffer[index])) * 4294967295);
} }
} }
break; break;
@ -1586,11 +1683,11 @@ static void stbir__encode_scanline(stbir__info* stbir_info, int num_pixels, void
for (n = 0; n < channels; n++) for (n = 0; n < channels; n++)
{ {
int index = pixel_index + n; int index = pixel_index + n;
((unsigned int*)output_buffer)[index] = (unsigned int)(floor(((double)stbir__linear_to_srgb(stbir__saturate(encode_buffer[index]))) * 4294967295 + 0.5f)); ((unsigned int*)output_buffer)[index] = (unsigned int)STBIR__ROUND_INT(((double)stbir__linear_to_srgb(stbir__saturate(encode_buffer[index]))) * 4294967295);
} }
if (!(stbir_info->flags&STBIR_FLAG_ALPHA_USES_COLORSPACE)) if (!(stbir_info->flags&STBIR_FLAG_ALPHA_USES_COLORSPACE))
((unsigned int*)output_buffer)[pixel_index + alpha_channel] = (unsigned int)(floor(((double)stbir__saturate(encode_buffer[pixel_index + alpha_channel])) * 4294967295 + 0.5f)); ((unsigned int*)output_buffer)[pixel_index + alpha_channel] = (unsigned int)STBIR__ROUND_INT(((double)stbir__saturate(encode_buffer[pixel_index + alpha_channel])) * 4294967295);
} }
break; break;
@ -1639,11 +1736,12 @@ static void stbir__resample_vertical_upsample(stbir__info* stbir_info, int n, in
int alpha_channel = stbir_info->alpha_channel; int alpha_channel = stbir_info->alpha_channel;
int type = stbir_info->type; int type = stbir_info->type;
int colorspace = stbir_info->colorspace; int colorspace = stbir_info->colorspace;
int kernel_pixel_width = stbir__get_filter_pixel_width_vertical(stbir_info); int kernel_pixel_width = stbir_info->vertical_filter_pixel_width;
void* output_data = stbir_info->output_data; void* output_data = stbir_info->output_data;
float* encode_buffer = stbir_info->encode_buffer; float* encode_buffer = stbir_info->encode_buffer;
int decode = STBIR__DECODE(type, colorspace); int decode = STBIR__DECODE(type, colorspace);
int coefficient_width = stbir__get_coefficient_width(stbir_info->vertical_filter, stbir_info->vertical_scale); int coefficient_width = stbir_info->vertical_coefficient_width;
int coefficient_counter;
int contributor = n; int contributor = n;
float* ring_buffer = stbir_info->ring_buffer; float* ring_buffer = stbir_info->ring_buffer;
@ -1664,22 +1762,82 @@ static void stbir__resample_vertical_upsample(stbir__info* stbir_info, int n, in
memset(encode_buffer, 0, output_w * sizeof(float) * channels); memset(encode_buffer, 0, output_w * sizeof(float) * channels);
for (x = 0; x < output_w; x++) coefficient_counter = 0;
{ switch (channels) {
int in_pixel_index = x * channels; case 1:
int coefficient_counter = 0;
for (k = n0; k <= n1; k++) for (k = n0; k <= n1; k++)
{ {
int coefficient_index = coefficient_counter++; int coefficient_index = coefficient_counter++;
float* ring_buffer_entry = stbir__get_ring_buffer_scanline(k, ring_buffer, ring_buffer_begin_index, ring_buffer_first_scanline, kernel_pixel_width, ring_buffer_length); float* ring_buffer_entry = stbir__get_ring_buffer_scanline(k, ring_buffer, ring_buffer_begin_index, ring_buffer_first_scanline, kernel_pixel_width, ring_buffer_length);
float coefficient = vertical_coefficients[coefficient_group + coefficient_index]; float coefficient = vertical_coefficients[coefficient_group + coefficient_index];
for (x = 0; x < output_w; x++)
{
int in_pixel_index = x * channels;
encode_buffer[in_pixel_index + 0] += ring_buffer_entry[in_pixel_index + 0] * coefficient;
}
}
break;
case 2:
for (k = n0; k <= n1; k++)
{
int coefficient_index = coefficient_counter++;
float* ring_buffer_entry = stbir__get_ring_buffer_scanline(k, ring_buffer, ring_buffer_begin_index, ring_buffer_first_scanline, kernel_pixel_width, ring_buffer_length);
float coefficient = vertical_coefficients[coefficient_group + coefficient_index];
for (x = 0; x < output_w; x++)
{
int in_pixel_index = x * channels;
encode_buffer[in_pixel_index + 0] += ring_buffer_entry[in_pixel_index + 0] * coefficient;
encode_buffer[in_pixel_index + 1] += ring_buffer_entry[in_pixel_index + 1] * coefficient;
}
}
break;
case 3:
for (k = n0; k <= n1; k++)
{
int coefficient_index = coefficient_counter++;
float* ring_buffer_entry = stbir__get_ring_buffer_scanline(k, ring_buffer, ring_buffer_begin_index, ring_buffer_first_scanline, kernel_pixel_width, ring_buffer_length);
float coefficient = vertical_coefficients[coefficient_group + coefficient_index];
for (x = 0; x < output_w; x++)
{
int in_pixel_index = x * channels;
encode_buffer[in_pixel_index + 0] += ring_buffer_entry[in_pixel_index + 0] * coefficient;
encode_buffer[in_pixel_index + 1] += ring_buffer_entry[in_pixel_index + 1] * coefficient;
encode_buffer[in_pixel_index + 2] += ring_buffer_entry[in_pixel_index + 2] * coefficient;
}
}
break;
case 4:
for (k = n0; k <= n1; k++)
{
int coefficient_index = coefficient_counter++;
float* ring_buffer_entry = stbir__get_ring_buffer_scanline(k, ring_buffer, ring_buffer_begin_index, ring_buffer_first_scanline, kernel_pixel_width, ring_buffer_length);
float coefficient = vertical_coefficients[coefficient_group + coefficient_index];
for (x = 0; x < output_w; x++)
{
int in_pixel_index = x * channels;
encode_buffer[in_pixel_index + 0] += ring_buffer_entry[in_pixel_index + 0] * coefficient;
encode_buffer[in_pixel_index + 1] += ring_buffer_entry[in_pixel_index + 1] * coefficient;
encode_buffer[in_pixel_index + 2] += ring_buffer_entry[in_pixel_index + 2] * coefficient;
encode_buffer[in_pixel_index + 3] += ring_buffer_entry[in_pixel_index + 3] * coefficient;
}
}
break;
default:
for (k = n0; k <= n1; k++)
{
int coefficient_index = coefficient_counter++;
float* ring_buffer_entry = stbir__get_ring_buffer_scanline(k, ring_buffer, ring_buffer_begin_index, ring_buffer_first_scanline, kernel_pixel_width, ring_buffer_length);
float coefficient = vertical_coefficients[coefficient_group + coefficient_index];
for (x = 0; x < output_w; x++)
{
int in_pixel_index = x * channels;
int c; int c;
for (c = 0; c < channels; c++) for (c = 0; c < channels; c++)
encode_buffer[in_pixel_index + c] += ring_buffer_entry[in_pixel_index + c] * coefficient; encode_buffer[in_pixel_index + c] += ring_buffer_entry[in_pixel_index + c] * coefficient;
} }
} }
break;
}
stbir__encode_scanline(stbir_info, output_w, (char *) output_data + output_row_start, encode_buffer, channels, alpha_channel, decode); stbir__encode_scanline(stbir_info, output_w, (char *) output_data + output_row_start, encode_buffer, channels, alpha_channel, decode);
} }
@ -1691,11 +1849,11 @@ static void stbir__resample_vertical_downsample(stbir__info* stbir_info, int n,
stbir__contributors* vertical_contributors = stbir_info->vertical_contributors; stbir__contributors* vertical_contributors = stbir_info->vertical_contributors;
float* vertical_coefficients = stbir_info->vertical_coefficients; float* vertical_coefficients = stbir_info->vertical_coefficients;
int channels = stbir_info->channels; int channels = stbir_info->channels;
int kernel_pixel_width = stbir__get_filter_pixel_width_vertical(stbir_info); int kernel_pixel_width = stbir_info->vertical_filter_pixel_width;
void* output_data = stbir_info->output_data; void* output_data = stbir_info->output_data;
float* horizontal_buffer = stbir_info->horizontal_buffer; float* horizontal_buffer = stbir_info->horizontal_buffer;
int coefficient_width = stbir__get_coefficient_width(stbir_info->vertical_filter, stbir_info->vertical_scale); int coefficient_width = stbir_info->vertical_coefficient_width;
int contributor = n + stbir__get_filter_pixel_margin_vertical(stbir_info); int contributor = n + stbir_info->vertical_filter_pixel_margin;
float* ring_buffer = stbir_info->ring_buffer; float* ring_buffer = stbir_info->ring_buffer;
int ring_buffer_begin_index = stbir_info->ring_buffer_begin_index; int ring_buffer_begin_index = stbir_info->ring_buffer_begin_index;
@ -1717,6 +1875,42 @@ static void stbir__resample_vertical_downsample(stbir__info* stbir_info, int n,
float* ring_buffer_entry = stbir__get_ring_buffer_scanline(k, ring_buffer, ring_buffer_begin_index, ring_buffer_first_scanline, kernel_pixel_width, ring_buffer_length); float* ring_buffer_entry = stbir__get_ring_buffer_scanline(k, ring_buffer, ring_buffer_begin_index, ring_buffer_first_scanline, kernel_pixel_width, ring_buffer_length);
switch (channels) {
case 1:
for (x = 0; x < output_w; x++)
{
int in_pixel_index = x * 1;
ring_buffer_entry[in_pixel_index + 0] += horizontal_buffer[in_pixel_index + 0] * coefficient;
}
break;
case 2:
for (x = 0; x < output_w; x++)
{
int in_pixel_index = x * 2;
ring_buffer_entry[in_pixel_index + 0] += horizontal_buffer[in_pixel_index + 0] * coefficient;
ring_buffer_entry[in_pixel_index + 1] += horizontal_buffer[in_pixel_index + 1] * coefficient;
}
break;
case 3:
for (x = 0; x < output_w; x++)
{
int in_pixel_index = x * 3;
ring_buffer_entry[in_pixel_index + 0] += horizontal_buffer[in_pixel_index + 0] * coefficient;
ring_buffer_entry[in_pixel_index + 1] += horizontal_buffer[in_pixel_index + 1] * coefficient;
ring_buffer_entry[in_pixel_index + 2] += horizontal_buffer[in_pixel_index + 2] * coefficient;
}
break;
case 4:
for (x = 0; x < output_w; x++)
{
int in_pixel_index = x * 4;
ring_buffer_entry[in_pixel_index + 0] += horizontal_buffer[in_pixel_index + 0] * coefficient;
ring_buffer_entry[in_pixel_index + 1] += horizontal_buffer[in_pixel_index + 1] * coefficient;
ring_buffer_entry[in_pixel_index + 2] += horizontal_buffer[in_pixel_index + 2] * coefficient;
ring_buffer_entry[in_pixel_index + 3] += horizontal_buffer[in_pixel_index + 3] * coefficient;
}
break;
default:
for (x = 0; x < output_w; x++) for (x = 0; x < output_w; x++)
{ {
int in_pixel_index = x * channels; int in_pixel_index = x * channels;
@ -1725,6 +1919,8 @@ static void stbir__resample_vertical_downsample(stbir__info* stbir_info, int n,
for (c = 0; c < channels; c++) for (c = 0; c < channels; c++)
ring_buffer_entry[in_pixel_index + c] += horizontal_buffer[in_pixel_index + c] * coefficient; ring_buffer_entry[in_pixel_index + c] += horizontal_buffer[in_pixel_index + c] * coefficient;
} }
break;
}
} }
} }
@ -1743,7 +1939,7 @@ static void stbir__buffer_loop_upsample(stbir__info* stbir_info)
stbir__calculate_sample_range_upsample(y, out_scanlines_radius, scale_ratio, stbir_info->vertical_shift, &in_first_scanline, &in_last_scanline, &in_center_of_out); stbir__calculate_sample_range_upsample(y, out_scanlines_radius, scale_ratio, stbir_info->vertical_shift, &in_first_scanline, &in_last_scanline, &in_center_of_out);
STBIR__DEBUG_ASSERT(in_last_scanline - in_first_scanline <= stbir__get_filter_pixel_width_vertical(stbir_info)); STBIR__DEBUG_ASSERT(in_last_scanline - in_first_scanline <= stbir_info->vertical_filter_pixel_width);
if (stbir_info->ring_buffer_begin_index >= 0) if (stbir_info->ring_buffer_begin_index >= 0)
{ {
@ -1762,7 +1958,7 @@ static void stbir__buffer_loop_upsample(stbir__info* stbir_info)
else else
{ {
stbir_info->ring_buffer_first_scanline++; stbir_info->ring_buffer_first_scanline++;
stbir_info->ring_buffer_begin_index = (stbir_info->ring_buffer_begin_index + 1) % stbir__get_filter_pixel_width_vertical(stbir_info); stbir_info->ring_buffer_begin_index = (stbir_info->ring_buffer_begin_index + 1) % stbir_info->vertical_filter_pixel_width;
} }
} }
} }
@ -1820,7 +2016,7 @@ static void stbir__empty_ring_buffer(stbir__info* stbir_info, int first_necessar
else else
{ {
stbir_info->ring_buffer_first_scanline++; stbir_info->ring_buffer_first_scanline++;
stbir_info->ring_buffer_begin_index = (stbir_info->ring_buffer_begin_index + 1) % stbir__get_filter_pixel_width_vertical(stbir_info); stbir_info->ring_buffer_begin_index = (stbir_info->ring_buffer_begin_index + 1) % stbir_info->vertical_filter_pixel_width;
} }
} }
} }
@ -1832,7 +2028,7 @@ static void stbir__buffer_loop_downsample(stbir__info* stbir_info)
float scale_ratio = stbir_info->vertical_scale; float scale_ratio = stbir_info->vertical_scale;
int output_h = stbir_info->output_h; int output_h = stbir_info->output_h;
float in_pixels_radius = stbir__filter_info_table[stbir_info->vertical_filter].support(scale_ratio) / scale_ratio; float in_pixels_radius = stbir__filter_info_table[stbir_info->vertical_filter].support(scale_ratio) / scale_ratio;
int pixel_margin = stbir__get_filter_pixel_margin_vertical(stbir_info); int pixel_margin = stbir_info->vertical_filter_pixel_margin;
int max_y = stbir_info->input_h + pixel_margin; int max_y = stbir_info->input_h + pixel_margin;
STBIR__DEBUG_ASSERT(!stbir__use_height_upsampling(stbir_info)); STBIR__DEBUG_ASSERT(!stbir__use_height_upsampling(stbir_info));
@ -1844,7 +2040,7 @@ static void stbir__buffer_loop_downsample(stbir__info* stbir_info)
stbir__calculate_sample_range_downsample(y, in_pixels_radius, scale_ratio, stbir_info->vertical_shift, &out_first_scanline, &out_last_scanline, &out_center_of_in); stbir__calculate_sample_range_downsample(y, in_pixels_radius, scale_ratio, stbir_info->vertical_shift, &out_first_scanline, &out_last_scanline, &out_center_of_in);
STBIR__DEBUG_ASSERT(out_last_scanline - out_first_scanline <= stbir__get_filter_pixel_width_vertical(stbir_info)); STBIR__DEBUG_ASSERT(out_last_scanline - out_first_scanline <= stbir_info->vertical_filter_pixel_width);
if (out_last_scanline < 0 || out_first_scanline >= output_h) if (out_last_scanline < 0 || out_first_scanline >= output_h)
continue; continue;
@ -1915,9 +2111,12 @@ static stbir_uint32 stbir__calculate_memory(stbir__info *info)
int pixel_margin = stbir__get_filter_pixel_margin(info->horizontal_filter, info->horizontal_scale); int pixel_margin = stbir__get_filter_pixel_margin(info->horizontal_filter, info->horizontal_scale);
int filter_height = stbir__get_filter_pixel_width(info->vertical_filter, info->vertical_scale); int filter_height = stbir__get_filter_pixel_width(info->vertical_filter, info->vertical_scale);
info->horizontal_contributors_size = stbir__get_horizontal_contributors(info) * sizeof(stbir__contributors); info->horizontal_num_contributors = stbir__get_contributors(info->horizontal_scale, info->horizontal_filter, info->input_w, info->output_w);
info->vertical_num_contributors = stbir__get_contributors(info->vertical_scale , info->vertical_filter , info->input_h, info->output_h);
info->horizontal_contributors_size = info->horizontal_num_contributors * sizeof(stbir__contributors);
info->horizontal_coefficients_size = stbir__get_total_horizontal_coefficients(info) * sizeof(float); info->horizontal_coefficients_size = stbir__get_total_horizontal_coefficients(info) * sizeof(float);
info->vertical_contributors_size = stbir__get_vertical_contributors(info) * sizeof(stbir__contributors); info->vertical_contributors_size = info->vertical_num_contributors * sizeof(stbir__contributors);
info->vertical_coefficients_size = stbir__get_total_vertical_coefficients(info) * sizeof(float); info->vertical_coefficients_size = stbir__get_total_vertical_coefficients(info) * sizeof(float);
info->decode_buffer_size = (info->input_w + pixel_margin * 2) * info->channels * sizeof(float); info->decode_buffer_size = (info->input_w + pixel_margin * 2) * info->channels * sizeof(float);
info->horizontal_buffer_size = info->output_w * info->channels * sizeof(float); info->horizontal_buffer_size = info->output_w * info->channels * sizeof(float);
@ -2018,8 +2217,15 @@ static int stbir__resize_allocated(stbir__info *info,
info->edge_vertical = edge_vertical; info->edge_vertical = edge_vertical;
info->colorspace = colorspace; info->colorspace = colorspace;
info->horizontal_coefficient_width = stbir__get_coefficient_width (info->horizontal_filter, info->horizontal_scale);
info->vertical_coefficient_width = stbir__get_coefficient_width (info->vertical_filter , info->vertical_scale );
info->horizontal_filter_pixel_width = stbir__get_filter_pixel_width (info->horizontal_filter, info->horizontal_scale);
info->vertical_filter_pixel_width = stbir__get_filter_pixel_width (info->vertical_filter , info->vertical_scale );
info->horizontal_filter_pixel_margin = stbir__get_filter_pixel_margin(info->horizontal_filter, info->horizontal_scale);
info->vertical_filter_pixel_margin = stbir__get_filter_pixel_margin(info->vertical_filter , info->vertical_scale );
info->ring_buffer_length_bytes = info->output_w * info->channels * sizeof(float); info->ring_buffer_length_bytes = info->output_w * info->channels * sizeof(float);
info->decode_buffer_pixels = info->input_w + stbir__get_filter_pixel_margin_horizontal(info) * 2; info->decode_buffer_pixels = info->input_w + info->horizontal_filter_pixel_margin * 2;
#define STBIR__NEXT_MEMPTR(current, newtype) (newtype*)(((unsigned char*)current) + current##_size) #define STBIR__NEXT_MEMPTR(current, newtype) (newtype*)(((unsigned char*)current) + current##_size)

51
tests/resample_test.cpp
View File

@ -141,6 +141,34 @@ static void resizer(int argc, char **argv)
exit(0); exit(0);
} }
static void performance(int argc, char **argv)
{
unsigned char* input_pixels;
unsigned char* output_pixels;
int w, h, count;
int n, i;
int out_w, out_h, srgb=1;
input_pixels = stbi_load(argv[1], &w, &h, &n, 0);
#if 1
out_w = w/4; out_h h/4; count=100; // 1
#elif 0
out_w = w*2; out_h = h/4; count=20; // 2 // note this is structured pessimily, would be much faster to downsample vertically first
#elif 0
out_w = w/4; out_h = h*2; count=50; // 3
#elif 0
out_w = w*3; out_h = h*3; count=5; srgb=0; // 4
#else
out_w = w*3; out_h = h*3; count=3; // 5 // this is dominated by linear->sRGB conversion
#endif
output_pixels = (unsigned char*) malloc(out_w*out_h*n);
for (i=0; i < count; ++i)
if (srgb)
stbir_resize_uint8_srgb(input_pixels, w, h, 0, output_pixels, out_w, out_h, 0, n, -1,0);
else
stbir_resize_uint8(input_pixels, w, h, 0, output_pixels, out_w, out_h, 0, n);
exit(0);
}
void test_suite(int argc, char **argv); void test_suite(int argc, char **argv);
@ -153,6 +181,7 @@ int main(int argc, char** argv)
int out_w, out_h, out_stride; int out_w, out_h, out_stride;
//resizer(argc, argv); //resizer(argc, argv);
//performance(argc, argv);
#if 1 #if 1
test_suite(argc, argv); test_suite(argc, argv);
@ -619,10 +648,10 @@ void verify_box(void)
+ image88[j*2+0][i*2+1] + image88[j*2+0][i*2+1]
+ image88[j*2+1][i*2+0] + image88[j*2+1][i*2+0]
+ image88[j*2+1][i*2+1]; + image88[j*2+1][i*2+1];
STBIR_ASSERT(output44[j][i] == ((n+2)>>2)); STBIR_ASSERT(output44[j][i] == ((n+2)>>2) || output44[j][i] == ((n+1)>>2)); // can't guarantee exact rounding due to numerical precision
t += n; t += n;
} }
STBIR_ASSERT(output11[0][0] == ((t+32)>>6)); STBIR_ASSERT(output11[0][0] == ((t+32)>>6) || output11[0][0] == ((t+31)>>6)); // can't guarantee exact rounding due to numerical precision
} }
void verify_filter_normalized(stbir_filter filter, int output_size, unsigned int value) void verify_filter_normalized(stbir_filter filter, int output_size, unsigned int value)
@ -646,6 +675,8 @@ void test_filters(void)
{ {
int i,j; int i,j;
mtsrand(0);
for (i=0; i < sizeof(image88); ++i) for (i=0; i < sizeof(image88); ++i)
image88[0][i] = mtrand() & 255; image88[0][i] = mtrand() & 255;
verify_box(); verify_box();
@ -673,25 +704,25 @@ void test_filters(void)
verify_filter_normalized(STBIR_FILTER_BOX, 8, value); verify_filter_normalized(STBIR_FILTER_BOX, 8, value);
verify_filter_normalized(STBIR_FILTER_TRIANGLE, 8, value); verify_filter_normalized(STBIR_FILTER_TRIANGLE, 8, value);
verify_filter_normalized(STBIR_FILTER_CUBIC, 8, value); verify_filter_normalized(STBIR_FILTER_CUBICBSPLINE, 8, value);
verify_filter_normalized(STBIR_FILTER_CATMULLROM, 8, value); verify_filter_normalized(STBIR_FILTER_CATMULLROM, 8, value);
verify_filter_normalized(STBIR_FILTER_MITCHELL, 8, value); verify_filter_normalized(STBIR_FILTER_MITCHELL, 8, value);
verify_filter_normalized(STBIR_FILTER_BOX, 4, value); verify_filter_normalized(STBIR_FILTER_BOX, 4, value);
verify_filter_normalized(STBIR_FILTER_TRIANGLE, 4, value); verify_filter_normalized(STBIR_FILTER_TRIANGLE, 4, value);
verify_filter_normalized(STBIR_FILTER_CUBIC, 4, value); verify_filter_normalized(STBIR_FILTER_CUBICBSPLINE, 4, value);
verify_filter_normalized(STBIR_FILTER_CATMULLROM, 4, value); verify_filter_normalized(STBIR_FILTER_CATMULLROM, 4, value);
verify_filter_normalized(STBIR_FILTER_MITCHELL, 4, value); verify_filter_normalized(STBIR_FILTER_MITCHELL, 4, value);
verify_filter_normalized(STBIR_FILTER_BOX, 2, value); verify_filter_normalized(STBIR_FILTER_BOX, 2, value);
verify_filter_normalized(STBIR_FILTER_TRIANGLE, 2, value); verify_filter_normalized(STBIR_FILTER_TRIANGLE, 2, value);
verify_filter_normalized(STBIR_FILTER_CUBIC, 2, value); verify_filter_normalized(STBIR_FILTER_CUBICBSPLINE, 2, value);
verify_filter_normalized(STBIR_FILTER_CATMULLROM, 2, value); verify_filter_normalized(STBIR_FILTER_CATMULLROM, 2, value);
verify_filter_normalized(STBIR_FILTER_MITCHELL, 2, value); verify_filter_normalized(STBIR_FILTER_MITCHELL, 2, value);
verify_filter_normalized(STBIR_FILTER_BOX, 1, value); verify_filter_normalized(STBIR_FILTER_BOX, 1, value);
verify_filter_normalized(STBIR_FILTER_TRIANGLE, 1, value); verify_filter_normalized(STBIR_FILTER_TRIANGLE, 1, value);
verify_filter_normalized(STBIR_FILTER_CUBIC, 1, value); verify_filter_normalized(STBIR_FILTER_CUBICBSPLINE, 1, value);
verify_filter_normalized(STBIR_FILTER_CATMULLROM, 1, value); verify_filter_normalized(STBIR_FILTER_CATMULLROM, 1, value);
verify_filter_normalized(STBIR_FILTER_MITCHELL, 1, value); verify_filter_normalized(STBIR_FILTER_MITCHELL, 1, value);
@ -917,14 +948,14 @@ void test_suite(int argc, char **argv)
// filter tests // filter tests
resize_image(barbara, 2, 2, STBIR_FILTER_BOX , STBIR_EDGE_CLAMP, STBIR_COLORSPACE_SRGB, "test-output/barbara-upsample-nearest.png"); resize_image(barbara, 2, 2, STBIR_FILTER_BOX , STBIR_EDGE_CLAMP, STBIR_COLORSPACE_SRGB, "test-output/barbara-upsample-nearest.png");
resize_image(barbara, 2, 2, STBIR_FILTER_TRIANGLE , STBIR_EDGE_CLAMP, STBIR_COLORSPACE_SRGB, "test-output/barbara-upsample-bilinear.png"); resize_image(barbara, 2, 2, STBIR_FILTER_TRIANGLE , STBIR_EDGE_CLAMP, STBIR_COLORSPACE_SRGB, "test-output/barbara-upsample-bilinear.png");
resize_image(barbara, 2, 2, STBIR_FILTER_CUBIC , STBIR_EDGE_CLAMP, STBIR_COLORSPACE_SRGB, "test-output/barbara-upsample-bicubic.png"); resize_image(barbara, 2, 2, STBIR_FILTER_CUBICBSPLINE, STBIR_EDGE_CLAMP, STBIR_COLORSPACE_SRGB, "test-output/barbara-upsample-bicubic.png");
resize_image(barbara, 2, 2, STBIR_FILTER_CATMULLROM, STBIR_EDGE_CLAMP, STBIR_COLORSPACE_SRGB, "test-output/barbara-upsample-catmullrom.png"); resize_image(barbara, 2, 2, STBIR_FILTER_CATMULLROM , STBIR_EDGE_CLAMP, STBIR_COLORSPACE_SRGB, "test-output/barbara-upsample-catmullrom.png");
resize_image(barbara, 2, 2, STBIR_FILTER_MITCHELL , STBIR_EDGE_CLAMP, STBIR_COLORSPACE_SRGB, "test-output/barbara-upsample-mitchell.png"); resize_image(barbara, 2, 2, STBIR_FILTER_MITCHELL , STBIR_EDGE_CLAMP, STBIR_COLORSPACE_SRGB, "test-output/barbara-upsample-mitchell.png");
resize_image(barbara, 0.5f, 0.5f, STBIR_FILTER_BOX , STBIR_EDGE_CLAMP, STBIR_COLORSPACE_SRGB, "test-output/barbara-downsample-nearest.png"); resize_image(barbara, 0.5f, 0.5f, STBIR_FILTER_BOX , STBIR_EDGE_CLAMP, STBIR_COLORSPACE_SRGB, "test-output/barbara-downsample-nearest.png");
resize_image(barbara, 0.5f, 0.5f, STBIR_FILTER_TRIANGLE , STBIR_EDGE_CLAMP, STBIR_COLORSPACE_SRGB, "test-output/barbara-downsample-bilinear.png"); resize_image(barbara, 0.5f, 0.5f, STBIR_FILTER_TRIANGLE , STBIR_EDGE_CLAMP, STBIR_COLORSPACE_SRGB, "test-output/barbara-downsample-bilinear.png");
resize_image(barbara, 0.5f, 0.5f, STBIR_FILTER_CUBIC , STBIR_EDGE_CLAMP, STBIR_COLORSPACE_SRGB, "test-output/barbara-downsample-bicubic.png"); resize_image(barbara, 0.5f, 0.5f, STBIR_FILTER_CUBICBSPLINE, STBIR_EDGE_CLAMP, STBIR_COLORSPACE_SRGB, "test-output/barbara-downsample-bicubic.png");
resize_image(barbara, 0.5f, 0.5f, STBIR_FILTER_CATMULLROM, STBIR_EDGE_CLAMP, STBIR_COLORSPACE_SRGB, "test-output/barbara-downsample-catmullrom.png"); resize_image(barbara, 0.5f, 0.5f, STBIR_FILTER_CATMULLROM , STBIR_EDGE_CLAMP, STBIR_COLORSPACE_SRGB, "test-output/barbara-downsample-catmullrom.png");
resize_image(barbara, 0.5f, 0.5f, STBIR_FILTER_MITCHELL , STBIR_EDGE_CLAMP, STBIR_COLORSPACE_SRGB, "test-output/barbara-downsample-mitchell.png"); resize_image(barbara, 0.5f, 0.5f, STBIR_FILTER_MITCHELL , STBIR_EDGE_CLAMP, STBIR_COLORSPACE_SRGB, "test-output/barbara-downsample-mitchell.png");
for (i = 10; i < 100; i++) for (i = 10; i < 100; i++)

5
tests/resize.dsp
View File

@ -39,9 +39,10 @@ RSC=rc.exe
# PROP Use_Debug_Libraries 0 # PROP Use_Debug_Libraries 0
# PROP Output_Dir "Release" # PROP Output_Dir "Release"
# PROP Intermediate_Dir "Release" # PROP Intermediate_Dir "Release"
# PROP Ignore_Export_Lib 0
# PROP Target_Dir "" # PROP Target_Dir ""
# ADD BASE CPP /nologo /W3 /GX /O2 /D "WIN32" /D "NDEBUG" /D "_CONSOLE" /D "_MBCS" /YX /FD /c # ADD BASE CPP /nologo /W3 /GX /O2 /D "WIN32" /D "NDEBUG" /D "_CONSOLE" /D "_MBCS" /YX /FD /c
# ADD CPP /nologo /W3 /GX /O2 /D "WIN32" /D "NDEBUG" /D "_CONSOLE" /D "_MBCS" /YX /FD /c # ADD CPP /nologo /G6 /W3 /GX /Z7 /O2 /I ".." /D "WIN32" /D "NDEBUG" /D "_CONSOLE" /D "_MBCS" /YX /FD /c
# ADD BASE RSC /l 0x409 /d "NDEBUG" # ADD BASE RSC /l 0x409 /d "NDEBUG"
# ADD RSC /l 0x409 /d "NDEBUG" # ADD RSC /l 0x409 /d "NDEBUG"
BSC32=bscmake.exe BSC32=bscmake.exe
@ -49,7 +50,7 @@ BSC32=bscmake.exe
# ADD BSC32 /nologo # ADD BSC32 /nologo
LINK32=link.exe LINK32=link.exe
# ADD BASE LINK32 kernel32.lib user32.lib gdi32.lib winspool.lib comdlg32.lib advapi32.lib shell32.lib ole32.lib oleaut32.lib uuid.lib odbc32.lib odbccp32.lib kernel32.lib user32.lib gdi32.lib winspool.lib comdlg32.lib advapi32.lib shell32.lib ole32.lib oleaut32.lib uuid.lib odbc32.lib odbccp32.lib /nologo /subsystem:console /machine:I386 # ADD BASE LINK32 kernel32.lib user32.lib gdi32.lib winspool.lib comdlg32.lib advapi32.lib shell32.lib ole32.lib oleaut32.lib uuid.lib odbc32.lib odbccp32.lib kernel32.lib user32.lib gdi32.lib winspool.lib comdlg32.lib advapi32.lib shell32.lib ole32.lib oleaut32.lib uuid.lib odbc32.lib odbccp32.lib /nologo /subsystem:console /machine:I386
# ADD LINK32 kernel32.lib user32.lib gdi32.lib winspool.lib comdlg32.lib advapi32.lib shell32.lib ole32.lib oleaut32.lib uuid.lib odbc32.lib odbccp32.lib kernel32.lib user32.lib gdi32.lib winspool.lib comdlg32.lib advapi32.lib shell32.lib ole32.lib oleaut32.lib uuid.lib odbc32.lib odbccp32.lib /nologo /subsystem:console /machine:I386 # ADD LINK32 kernel32.lib user32.lib gdi32.lib winspool.lib comdlg32.lib advapi32.lib shell32.lib ole32.lib oleaut32.lib uuid.lib odbc32.lib odbccp32.lib kernel32.lib user32.lib gdi32.lib winspool.lib comdlg32.lib advapi32.lib shell32.lib ole32.lib oleaut32.lib uuid.lib odbc32.lib odbccp32.lib /nologo /subsystem:console /debug /machine:I386
!ELSEIF "$(CFG)" == "resize - Win32 Debug" !ELSEIF "$(CFG)" == "resize - Win32 Debug"

3
tools/README.list
View File

@ -2,11 +2,12 @@ stb_vorbis.c | audio | decode ogg vorbis files from file/memory to float
stb_image.h | graphics | image loading/decoding from file/memory: JPG, PNG, TGA, BMP, PSD, GIF, HDR, PIC stb_image.h | graphics | image loading/decoding from file/memory: JPG, PNG, TGA, BMP, PSD, GIF, HDR, PIC
stb_truetype.h | graphics | parse, decode, and rasterize characters from truetype fonts stb_truetype.h | graphics | parse, decode, and rasterize characters from truetype fonts
stb_image_write.h | graphics | image writing to disk: PNG, TGA, BMP stb_image_write.h | graphics | image writing to disk: PNG, TGA, BMP
stb_image_resize.h | graphics | resize images larger/smaller with good quality
stretchy_buffer.h | utility | typesafe dynamic array for C (i.e. approximation to vector<>), doesn't compile as C++ stretchy_buffer.h | utility | typesafe dynamic array for C (i.e. approximation to vector<>), doesn't compile as C++
stb_textedit.h | UI | guts of a text editor for games etc implementing them from scratch stb_textedit.h | UI | guts of a text editor for games etc implementing them from scratch
stb_dxt.h | 3D graphics | Fabian "ryg" Giesen's real-time DXT compressor stb_dxt.h | 3D graphics | Fabian "ryg" Giesen's real-time DXT compressor
stb_herringbone_wang_tile.h | games | herringbone Wang tile map generator
stb_perlin.h | 3D graphics | revised Perlin noise (3D input, 1D output) stb_perlin.h | 3D graphics | revised Perlin noise (3D input, 1D output)
stb_herringbone_wang_tile.h | games | herringbone Wang tile map generator
stb_c_lexer.h | parsing | simplify writing parsers for C-like languages stb_c_lexer.h | parsing | simplify writing parsers for C-like languages
stb_divide.h | math | more useful 32-bit modulus e.g. "euclidean divide" stb_divide.h | math | more useful 32-bit modulus e.g. "euclidean divide"
stb.h | misc | helper functions for C, mostly redundant in C++; basically author's personal stuff stb.h | misc | helper functions for C, mostly redundant in C++; basically author's personal stuff

4
tools/make_readme.dsp
View File

@ -84,5 +84,9 @@ LINK32=link.exe
SOURCE=.\make_readme.c SOURCE=.\make_readme.c
# End Source File # End Source File
# Begin Source File
SOURCE=.\README.list
# End Source File
# End Target # End Target
# End Project # End Project