Add ssim perceptual downscaling

This commit is contained in:
Hybrid Dog 2018-02-28 19:56:07 +01:00 committed by est31
parent 395bc9af19
commit ff970a8687
3 changed files with 335 additions and 0 deletions

1
.gitignore vendored
View File

@ -1,3 +1,4 @@
www/tiles/ www/tiles/
www/tiles-* www/tiles-*
www/conf.json www/conf.json
downscaling/perc

1
downscaling/compile.sh Normal file
View File

@ -0,0 +1 @@
gcc -Wall -Ofast -march=native percept_down.c -o perc -lpng -lm

333
downscaling/percept_down.c Normal file
View File

@ -0,0 +1,333 @@
// this is two times as fast when compiled with -Ofast
// for not tileable textures
// see https://graphics.ethz.ch/~cengizo/Files/Sig15PerceptualDownscaling.pdf
#include <stdlib.h> // malloc, EXIT_*
#include <string.h> // memset
#include <math.h>
#include <png.h>
#define SQR_NP 2 // squareroot of the patch size, recommended: 2
#define EXIT_PNG(F) if (!F) { \
fprintf(stderr, "%s\n", bild.message); \
return EXIT_FAILURE; \
}
#define CLAMP(V, A, B) (V) < (A) ? (A) : (V) > (B) ? (B) : (V)
#define u8 unsigned char
struct pixel {
u8 r;
u8 g;
u8 b;
u8 a;
};
#define PIXELBYTES 4
struct matrix {
int w;
int h;
float *data;
};
struct image {
int w;
int h;
struct pixel *pixels;
};
/*! \brief get y, cb and cr values each in [0;1] from u8 r, g and b values
*
* there's gamma correction,
* see http://www.ericbrasseur.org/gamma.html?i=1#Assume_a_gamma_of_2.2
* 0.5 is added to cb and cr to have them in [0;1]
*/
static void rgb2ycbcr(u8 or, u8 og, u8 ob, float *y, float *cb, float *cr)
{
float divider = 1.0f / 255.0f;
float r = powf(or * divider, 2.2f);
float g = powf(og * divider, 2.2f);
float b = powf(ob * divider, 2.2f);
*y = (0.299f * r + 0.587f * g + 0.114f * b);
*cb = (-0.168736f * r - 0.331264f * g + 0.5f * b) + 0.5f;
*cr = (0.5f * r - 0.418688f * g - 0.081312f * b) + 0.5f;
}
/*! \brief the inverse of the function above
*
* numbers from http://www.equasys.de/colorconversion.html
* if values are too big or small, they're clamped
*/
static void ycbcr2rgb(float y, float cb, float cr, u8 *r, u8 *g, u8 *b)
{
float vr = (y + 1.402f * (cr - 0.5f));
float vg = (y - 0.344136f * (cb - 0.5f) - 0.714136f * (cr - 0.5f));
float vb = (y + 1.772f * (cb - 0.5f));
float exponent = 1.0f / 2.2f;
vr = powf(vr, exponent);
vg = powf(vg, exponent);
vb = powf(vb, exponent);
*r = CLAMP(vr * 255.0f, 0, 255);
*g = CLAMP(vg * 255.0f, 0, 255);
*b = CLAMP(vb * 255.0f, 0, 255);
}
/*! \brief Convert an rgba image to 4 ycbcr matrices with values in [0, 1]
*/
static struct matrix *image_to_matrices(struct image *bild)
{
int w = bild->w;
int h = bild->h;
struct matrix *matrices = malloc(
PIXELBYTES * sizeof(struct matrix));
for (int i = 0; i < PIXELBYTES; ++i) {
matrices[i].w = w;
matrices[i].h = h;
matrices[i].data = malloc(w * h * sizeof(float));
}
for (int i = 0; i < w * h; ++i) {
struct pixel px = bild->pixels[i];
// put y, cb, cr and transpatency into the matrices
rgb2ycbcr(px.r, px.g, px.b,
&matrices[0].data[i], &matrices[1].data[i], &matrices[2].data[i]);
float divider = 1.0f / 255.0f;
matrices[3].data[i] = px.a * divider;
}
return matrices;
}
/*! \brief Convert 4 matrices to an rgba image
*
* Note that matrices becomes freed.
*/
static struct image *matrices_to_image(struct matrix *matrices)
{
struct image *bild = malloc(sizeof(struct image));
int w = matrices[0].w;
int h = matrices[0].h;
bild->w = w;
bild->h = h;
struct pixel *pixels = malloc(w * h * PIXELBYTES);
for (int i = 0; i < w * h; ++i) {
struct pixel *px = &pixels[i];
ycbcr2rgb(matrices[0].data[i], matrices[1].data[i], matrices[2].data[i],
&px->r, &px->g, &px->b);
float a = matrices[3].data[i] * 255;
px->a = CLAMP(a, 0, 255);
}
for (int i = 0; i < PIXELBYTES; ++i) {
free(matrices[i].data);
}
free(matrices);
bild->pixels = pixels;
return bild;
}
/*! \brief The actual downscaling algorithm
*
* \param mat The 4 matrices obtained form image_to_matrices.
* \param s The factor by which the image should become downscaled.
*/
static void downscale_perc(struct matrix *mat, int s)
{
// preparation
int w = mat->w; // input width
int h = mat->h;
float *input = mat->data;
int w2 = w / s; // output width
int h2 = h / s;
int input_size = w * h * sizeof(float);
int output_size = input_size / (s * s);
//~ fprintf(stderr, "w, h, s: %d, %d, %d\n", w,h,s);
float *l = malloc(output_size);
float *l2 = malloc(output_size);
float *d = malloc(output_size);
// set d's entries to 0 (because it's used for a sum)
for (int i = 0; i < w2 * h2; ++i)
d[i] = 0;
// get l and l2, the input image and it's size are used only here
for (int ysm = 0; ysm < h2; ++ysm) {
for (int xsm = 0; xsm < w2; ++xsm) {
// xsm and ysm are coords for the subsampled image
int x = xsm * s;
int y = ysm * s;
float acc = 0;
float acc2 = 0;
for (int yc = y; yc < y + s; ++yc) {
for (int xc = x; xc < x + s; ++xc) {
// if xc or yc is out of the image, set it to the border
int y_cl = CLAMP(yc, 0, h-1);
int x_cl = CLAMP(xc, 0, w-1);
float v = input[y_cl * w + x_cl];
acc += v;
acc2 += v * v;
}
}
int ism = ysm*w2+xsm;
float divider = 1.0f / (s * s);
l[ism] = acc * divider;
l2[ism] = acc2 * divider;
}
}
float patch_sz_div = 1.0f / (SQR_NP * SQR_NP);
// calculate the average of the results of all possible patch sets
for (int y_offset = 0; y_offset > -SQR_NP; --y_offset) {
for (int x_offset = 0; x_offset > -SQR_NP; --x_offset) {
float *m = malloc(output_size);
float *r = malloc(output_size);
// get m
for (int y = 0; y < h2; ++y) {
for (int x = 0; x < w2; ++x) {
float acc = 0;
int ys = y - (y + SQR_NP + y_offset) % SQR_NP;
int xs = x - (x + SQR_NP + x_offset) % SQR_NP;
for (int yc = ys; yc < ys + SQR_NP; ++yc) {
for (int xc = xs; xc < xs + SQR_NP; ++xc) {
int y_cl = CLAMP(yc, 0, h2-1);
int x_cl = CLAMP(xc, 0, w2-1);
int i = y_cl * w2 + x_cl;
acc += l[i];
}
}
m[y*w2+x] = acc * patch_sz_div;
}
}
// get r
for (int y = 0; y < h2; ++y) {
for (int x = 0; x < w2; ++x) {
float acc = 0;
float acc2 = 0;
int ys = y - (y + SQR_NP + y_offset) % SQR_NP;
int xs = x - (x + SQR_NP + x_offset) % SQR_NP;
for (int yc = ys; yc < ys + SQR_NP; ++yc) {
for (int xc = xs; xc < xs + SQR_NP; ++xc) {
int y_cl = CLAMP(yc, 0, h2-1);
int x_cl = CLAMP(xc, 0, w2-1);
int i = y_cl * w2 + x_cl;
acc += l[i] * l[i];
acc2 += l2[i];
}
}
int i = y*w2+x;
float mv = m[i];
float slv = acc * patch_sz_div - mv * mv;
float shv = acc2 * patch_sz_div - mv * mv;
if (slv >= 0.000001f) // epsilon is 10⁻⁶
r[i] = sqrtf(shv / slv);
else
r[i] = 0;
}
}
// get d, which is the output
for (int y = 0; y < h2; ++y) {
for (int x = 0; x < w2; ++x) {
float acc_m = 0;
float acc_r = 0;
float acc_t = 0;
int ys = y - (y + SQR_NP + y_offset) % SQR_NP;
int xs = x - (x + SQR_NP + x_offset) % SQR_NP;
for (int yc = ys; yc < ys + SQR_NP; ++yc) {
for (int xc = xs; xc < xs + SQR_NP; ++xc) {
int y_cl = CLAMP(yc, 0, h2-1);
int x_cl = CLAMP(xc, 0, w2-1);
int i = y_cl * w2 + x_cl;
acc_m += m[i];
acc_r += r[i];
acc_t += r[i] * m[i];
}
}
int i = y*w2+x;
d[i] += (
acc_m * patch_sz_div
+ acc_r * patch_sz_div * l[i]
- acc_t * patch_sz_div
);
}
}
free(m);
free(r);
}
}
// divide values in d for the (arithmetic) average
for (int i = 0; i < w2 * h2; ++i)
d[i] *= patch_sz_div;
// update the matrix
mat->data = d;
mat->w = w2;
mat->h = h2;
// tidy up
free(input);
free(l);
free(l2);
}
/*! \brief Function which calls functions for downscaling
*
* \param bild The image, it's content is changed when finished.
* \param downscale_factor Must be a natural number.
*/
void downscale_an_image(struct image **bild, int downscale_factor)
{
struct matrix *matrices = image_to_matrices(*bild);
for (int i = 0; i < PIXELBYTES; ++i) {
downscale_perc(&(matrices[i]), downscale_factor);
}
*bild = matrices_to_image(matrices);
}
int main(int argc, char **args)
{
if (argc != 2) {
fprintf(stderr, "Missing arguments, usage: ./perc <downscaling_factor>"
"\n");
return EXIT_FAILURE;
}
int downscaling_factor = atoi(args[1]);
if (downscaling_factor < 2) {
fprintf(stderr, "Invalid downscaling factor: %d\n",
downscaling_factor);
return EXIT_FAILURE;
}
png_image bild;
memset(&bild, 0, sizeof(bild));
bild.version = PNG_IMAGE_VERSION;
EXIT_PNG(png_image_begin_read_from_stdio(&bild, stdin))
int w = bild.width;
int h = bild.height;
bild.flags = PNG_IMAGE_FLAG_COLORSPACE_NOT_sRGB;
bild.format = PNG_FORMAT_RGBA;
struct pixel *pixels = malloc(w * h * 4);
EXIT_PNG(png_image_finish_read(&bild, NULL, pixels, 0, NULL))
struct image origpic = {w = w, h = h, pixels = pixels};
struct image *newpic = &origpic;
downscale_an_image(&newpic, downscaling_factor);
bild.width = newpic->w;
bild.height = newpic->h;
free(pixels);
pixels = newpic->pixels;
free(newpic);
EXIT_PNG(png_image_write_to_stdio(&bild, stdout, 0, pixels, 0, NULL));
free(pixels); // redundant free to feed valgrind
return EXIT_SUCCESS;
}