Replaced libvpx by libwebp (first draft; Windows only)

ext/gd/libgd/gd_webp.c has been taken from libgd[1]. Mainly, gd_error(X) has
been subsituted by zend_error(E_ERROR, X) and BGD_DECLARE(X) by X. Further
modifications are obvious from the diff.

All GD tests are passing, what raises hope, but we need more WebP tests,
anyway.

[1] <7ec030c4f1/src/gd_webp.c>
This commit is contained in:
Christoph M. Becker 2015-07-14 00:29:05 +02:00 committed by Anatol Belski
parent 9b9448a54a
commit 00ba7e6215
4 changed files with 110 additions and 1227 deletions

View File

@ -2,7 +2,7 @@
// vim:ft=javascript
ARG_WITH("gd", "Bundled GD support", "yes,shared");
ARG_WITH("libvpx", "vpx support", "yes");
ARG_WITH("libwebp", "webp support", "yes");
if (PHP_GD != "no") {
if (
@ -21,12 +21,13 @@ if (PHP_GD != "no") {
CHECK_HEADER_ADD_INCLUDE("xpm.h", "CFLAGS_GD", PHP_GD + ";" + PHP_PHP_BUILD + "\\include\\X11")
) {
if (PHP_LIBVPX != "no") {
if (CHECK_LIB("vpxmt.lib", "gd", PHP_GD) &&
CHECK_HEADER_ADD_INCLUDE("vp8.h", "CFLAGS_GD", PHP_GD + ";" + PHP_PHP_BUILD + "\\include\\vpx")) {
ADD_FLAG("CFLAGS_GD", "/D HAVE_LIBVPX /D HAVE_GD_WEBP");
if (PHP_LIBWEBP != "no") {
if (CHECK_LIB("libwebp.lib", "gd", PHP_GD) &&
CHECK_HEADER_ADD_INCLUDE("decode.h", "CFLAGS_GD", PHP_GD + ";" + PHP_PHP_BUILD + "\\include\\webp") &&
CHECK_HEADER_ADD_INCLUDE("encode.h", "CFLAGS_GD", PHP_GD + ";" + PHP_PHP_BUILD + "\\include\\webp")) {
ADD_FLAG("CFLAGS_GD", "/D HAVE_LIBWEBP /D HAVE_GD_WEBP");
} else {
WARNING("libvpx not enabled; libraries and headers not found");
WARNING("libwebp not enabled; libraries and headers not found");
}
}
CHECK_LIB("User32.lib", "gd", PHP_GD);
@ -38,7 +39,7 @@ if (PHP_GD != "no") {
gdft.c gd_gd2.c gd_gd.c gd_gif_in.c gd_gif_out.c gdhelpers.c gd_io.c gd_io_dp.c \
gd_io_file.c gd_io_ss.c gd_jpeg.c gdkanji.c gd_png.c gd_ss.c \
gdtables.c gd_topal.c gd_wbmp.c gdxpm.c wbmp.c xbm.c gd_security.c gd_transform.c \
gd_filter.c gd_pixelate.c gd_arc.c gd_rotate.c gd_color.c webpimg.c gd_webp.c \
gd_filter.c gd_pixelate.c gd_arc.c gd_rotate.c gd_color.c gd_webp.c \
gd_crop.c gd_interpolation.c gd_matrix.c", "gd");
AC_DEFINE('HAVE_LIBGD', 1, 'GD support');
ADD_FLAG("CFLAGS_GD", " \

View File

@ -1,28 +1,14 @@
#ifdef HAVE_LIBWEBP
#include <stdio.h>
#include <math.h>
#include <string.h>
#include <stdlib.h>
#include "gd.h"
#ifdef HAVE_LIBVPX
#include "webpimg.h"
#include "gdhelpers.h"
#include "webp/decode.h"
#include "webp/encode.h"
extern void gd_YUV420toRGBA(uint8* Y,
uint8* U,
uint8* V,
gdImagePtr im);
extern void gd_RGBAToYUV420(gdImagePtr im2,
uint8* Y,
uint8* U,
uint8* V);
const char * gdWebpGetVersionString()
{
return "not defined";
}
#define GD_WEBP_ALLOC_STEP (4*1024)
gdImagePtr gdImageCreateFromWebp (FILE * inFile)
{
@ -34,42 +20,28 @@ gdImagePtr gdImageCreateFromWebp (FILE * inFile)
return im;
}
gdImagePtr gdImageCreateFromWebpPtr (int size, void *data)
{
int width, height, ret;
unsigned char *Y = NULL;
unsigned char *U = NULL;
unsigned char *V = NULL;
gdImagePtr im;
ret = WebPDecode(data, size, &Y, &U, &V, &width, &height);
if (ret != webp_success) {
if (Y) free(Y);
if (U) free(U);
if (V) free(V);
php_gd_error("WebP decode: fail to decode input data");
return NULL;
}
im = gdImageCreateTrueColor(width, height);
if (!im) {
return NULL;
}
gd_YUV420toRGBA(Y, U, V, im);
gdIOCtx *in = gdNewDynamicCtxEx(size, data, 0);
if (!in)
return 0;
im = gdImageCreateFromWebpCtx(in);
in->gd_free(in);
return im;
}
#define GD_WEBP_ALLOC_STEP (4*1024)
gdImagePtr gdImageCreateFromWebpCtx (gdIOCtx * infile)
{
int width, height, ret;
unsigned char *filedata = NULL;
int width, height;
uint8_t *filedata = NULL;
uint8_t *argb = NULL;
unsigned char *read, *temp;
unsigned char *Y = NULL;
unsigned char *U = NULL;
unsigned char *V = NULL;
size_t size = 0, n;
gdImagePtr im;
int x, y;
uint8_t *p;
do {
temp = gdRealloc(filedata, size+GD_WEBP_ALLOC_STEP);
@ -80,31 +52,106 @@ gdImagePtr gdImageCreateFromWebpCtx (gdIOCtx * infile)
if (filedata) {
gdFree(filedata);
}
php_gd_error("WebP decode: realloc failed");
zend_error(E_ERROR, "WebP decode: realloc failed");
return NULL;
}
n = gdGetBuf(read, GD_WEBP_ALLOC_STEP, infile);
/* differs from upstream where gdGetBuf return 0 instead of EOF */
if (n>0 && n!=EOF) {
size += n;
}
} while (n>0 && n!=EOF);
ret = WebPDecode(filedata, size, &Y, &U, &V, &width, &height);
gdFree(filedata);
if (ret != webp_success) {
if (Y) free(Y);
if (U) free(U);
if (V) free(V);
php_gd_error("WebP decode: fail to decode input data");
if (WebPGetInfo(filedata,size, &width, &height) == 0) {
zend_error(E_ERROR, "gd-webp cannot get webp info");
gdFree(temp);
return NULL;
}
im = gdImageCreateTrueColor(width, height);
gd_YUV420toRGBA(Y, U, V, im);
if (!im) {
gdFree(temp);
return NULL;
}
argb = WebPDecodeARGB(filedata, size, &width, &height);
if (!argb) {
zend_error(E_ERROR, "gd-webp cannot allocate temporary buffer");
gdFree(temp);
gdImageDestroy(im);
return NULL;
}
for (y = 0, p = argb; y < height; y++) {
for (x = 0; x < width; x++) {
register uint8_t a = gdAlphaMax - (*(p++) >> 1);
register uint8_t r = *(p++);
register uint8_t g = *(p++);
register uint8_t b = *(p++);
im->tpixels[y][x] = gdTrueColorAlpha(r, g, b, a);
}
}
gdFree(filedata);
/* do not use gdFree here, in case gdFree/alloc is mapped to something else than libc */
free(argb);
gdFree(temp);
im->saveAlphaFlag = 1;
return im;
}
void gdImageWebpCtx (gdImagePtr im, gdIOCtx * outfile, int quantization)
{
uint8_t *argb;
int x, y;
uint8_t *p;
uint8_t *out;
size_t out_size;
if (im == NULL) {
return;
}
if (!gdImageTrueColor(im)) {
zend_error(E_ERROR, "Paletter image not supported by webp");
return;
}
if (quantization == -1) {
quantization = 80;
}
argb = (uint8_t *)gdMalloc(gdImageSX(im) * 4 * gdImageSY(im));
if (!argb) {
return;
}
p = argb;
for (y = 0; y < gdImageSY(im); y++) {
for (x = 0; x < gdImageSX(im); x++) {
register int c;
register char a;
c = im->tpixels[y][x];
a = gdTrueColorGetAlpha(c);
if (a == 127) {
a = 0;
} else {
a = 255 - ((a << 1) + (a >> 6));
}
*(p++) = gdTrueColorGetRed(c);
*(p++) = gdTrueColorGetGreen(c);
*(p++) = gdTrueColorGetBlue(c);
*(p++) = a;
}
}
out_size = WebPEncodeRGBA(argb, gdImageSX(im), gdImageSY(im), gdImageSX(im) * 4, quantization, &out);
if (out_size == 0) {
zend_error(E_ERROR, "gd-webp encoding failed");
goto freeargb;
}
gdPutBuf(out, out_size, outfile);
free(out);
freeargb:
gdFree(argb);
}
void gdImageWebpEx (gdImagePtr im, FILE * outFile, int quantization)
{
gdIOCtx *out = gdNewFileCtx(outFile);
@ -115,7 +162,7 @@ void gdImageWebpEx (gdImagePtr im, FILE * outFile, int quantization)
void gdImageWebp (gdImagePtr im, FILE * outFile)
{
gdIOCtx *out = gdNewFileCtx(outFile);
gdImageWebpCtx(im, out, -1);
gdImageWebpCtx(im, out, -1);
out->gd_free(out);
}
@ -139,74 +186,4 @@ void * gdImageWebpPtrEx (gdImagePtr im, int *size, int quantization)
out->gd_free(out);
return rv;
}
/*
* Maps normalized QP (quality) to VP8 QP
*/
int mapQualityToVP8QP(int quality) {
#define MIN_QUALITY 0
#define MAX_QUALITY 100
#define MIN_VP8QP 1
#define MAX_VP8QP 63
const float scale = MAX_VP8QP - MIN_VP8QP;
const float vp8qp =
scale * (MAX_QUALITY - quality) / (MAX_QUALITY - MIN_QUALITY) + MIN_VP8QP;
if (quality < MIN_QUALITY || quality > MAX_QUALITY) {
php_gd_error("Wrong quality value %d.", quality);
return -1;
}
return (int)(vp8qp + 0.5);
}
/* This routine is based in part on code from Dale Lutz (Safe Software Inc.)
* and in part on demo code from Chapter 15 of "PNG: The Definitive Guide"
* (http://www.cdrom.com/pub/png/pngbook.html).
*/
void gdImageWebpCtx (gdImagePtr im, gdIOCtx * outfile, int quantization)
{
int width = im->sx;
int height = im->sy;
int colors = im->colorsTotal;
int *open = im->open;
int yuv_width, yuv_height, yuv_nbytes, ret;
int vp8_quality;
unsigned char *Y = NULL,
*U = NULL,
*V = NULL;
unsigned char *filedata = NULL;
/* Conversion to Y,U,V buffer */
yuv_width = (width + 1) >> 1;
yuv_height = (height + 1) >> 1;
yuv_nbytes = width * height + 2 * yuv_width * yuv_height;
if ((Y = (unsigned char *)gdCalloc(yuv_nbytes, sizeof(unsigned char))) == NULL) {
php_gd_error("gd-webp error: cannot allocate Y buffer");
return;
}
vp8_quality = mapQualityToVP8QP(quantization);
U = Y + width * height;
V = U + yuv_width * yuv_height;
gd_RGBAToYUV420(im, Y, U, V);
/* Encode Y,U,V and write data to file */
ret = WebPEncode(Y, U, V, width, height, width, yuv_width, yuv_height, yuv_width,
vp8_quality, &filedata, &yuv_nbytes, NULL);
gdFree(Y);
if (ret != webp_success) {
if (filedata) {
free(filedata);
}
php_gd_error("gd-webp error: WebP Encoder failed");
return;
}
gdPutBuf (filedata, yuv_nbytes, outfile);
free(filedata);
}
#endif /* HAVE_LIBVPX */
#endif /* HAVE_LIBWEBP */

View File

@ -1,914 +0,0 @@
/*===========================================================================*
- Copyright 2010 Google Inc.
-
- This code is licensed under the same terms as WebM:
- Software License Agreement: http://www.webmproject.org/license/software/
- Additional IP Rights Grant: http://www.webmproject.org/license/additional/
*===========================================================================*/
/*
* Encoding/Decoding of WebP still image compression format.
*
* 1. WebPDecode: Takes an array of bytes (string) corresponding to the WebP
* encoded image and generates output in the YUV format with
* the color components U, V subsampled to 1/2 resolution along
* each dimension.
*
* 2. YUV420toRGBA: Converts from YUV (with color subsampling) such as produced
* by the WebPDecode routine into 32 bits per pixel RGBA data
* array. This data array can be directly used by the Leptonica
* Pix in-memory image format.
*
* 3. WebPEncode: Takes a Y, U, V data buffers (with color components U and V
* subsampled to 1/2 resolution) and generates the WebP string
*
* 4. RGBAToYUV420: Generates Y, U, V data (with color subsampling) from 32 bits
* per pixel RGBA data buffer. The resulting YUV data can be
* directly fed into the WebPEncode routine.
*
* 5. AdjustColorspace:
*
* 6. AdjustColorspaceBack:
*/
#include "gd.h"
#ifdef HAVE_LIBVPX
#include "webpimg.h"
#include <math.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <sys/stat.h>
#include "vpx/vpx_decoder.h"
#include "vpx/vp8dx.h"
#include "vpx/vpx_encoder.h"
#include "vpx/vp8cx.h"
#include "gd.h"
/*---------------------------------------------------------------------*
* color conversions *
*---------------------------------------------------------------------*/
#ifndef inline
# define inline __inline
#endif
static inline int clip(float v, int a, int b) {
return (v > b) ? b : (v < 0) ? 0 : (int)(v);
}
enum {
COLOR_RED = 1,
COLOR_GREEN = 2,
COLOR_BLUE = 3,
ALPHA_CHANNEL = 0
};
/* endian neutral extractions of ARGB from a 32 bit pixel */
static const uint32 RED_SHIFT =
8 * (sizeof(uint32) - 1 - COLOR_RED); /* 16 */
static const uint32 GREEN_SHIFT =
8 * (sizeof(uint32) - 1 - COLOR_GREEN); /* 8 */
static const uint32 BLUE_SHIFT =
8 * (sizeof(uint32) - 1 - COLOR_BLUE); /* 0 */
static const uint32 ALPHA_SHIFT =
8 * (sizeof(uint32) - 1 - ALPHA_CHANNEL); /* 24 */
static inline int GetRed(const uint32* rgba) {
return gdTrueColorGetRed(*rgba);
}
static inline int GetGreen(const uint32* rgba) {
return gdTrueColorGetGreen(*rgba);
}
static inline int GetBlue(const uint32* rgba) {
return gdTrueColorGetBlue(*rgba);
}
enum { YUV_FRAC = 16 };
static inline int clip_uv(int v) {
v = (v + (257 << (YUV_FRAC + 2 - 1))) >> (YUV_FRAC + 2);
return ((v & ~0xff) == 0) ? v : v < 0 ? 0u : 255u;
}
/* YUV <-----> RGB conversions */
/* The exact naming is Y'CbCr, following the ITU-R BT.601 standard.
* More information at: http://en.wikipedia.org/wiki/YCbCr
*/
static inline int GetLumaY(int r, int g, int b) {
const int kRound = (1 << (YUV_FRAC - 1)) + (16 << YUV_FRAC);
// Y = 0.2569 * R + 0.5044 * G + 0.0979 * B + 16
const int luma = 16839 * r + 33059 * g + 6420 * b;
return (luma + kRound) >> YUV_FRAC;
}
static inline int GetLumaYfromPtr(uint32* rgba) {
const int r = GetRed(rgba);
const int g = GetGreen(rgba);
const int b = GetBlue(rgba);
return GetLumaY(r, g, b);
}
static inline int GetChromaU(int r, int g, int b) {
// U = -0.1483 * R - 0.2911 * G + 0.4394 * B + 128
return clip_uv(-9719 * r - 19081 * g + 28800 * b);
}
static inline int GetChromaV(int r, int g, int b) {
// V = 0.4394 * R - 0.3679 * G - 0.0715 * B + 128
return clip_uv(+28800 * r - 24116 * g - 4684 * b);
}
/* Converts YUV to RGB and writes into a 32 bit pixel in endian
* neutral fashion
*/
enum { RGB_FRAC = 16, RGB_HALF = (1 << RGB_FRAC) / 2,
RGB_RANGE_MIN = -227, RGB_RANGE_MAX = 256 + 226 };
static int init_done = 0;
static int16_t kVToR[256], kUToB[256];
static int32_t kVToG[256], kUToG[256];
static uint8_t kClip[RGB_RANGE_MAX - RGB_RANGE_MIN];
static void InitTables() {
int i;
for (i = 0; i < 256; ++i) {
kVToR[i] = (89858 * (i - 128) + RGB_HALF) >> RGB_FRAC;
kUToG[i] = -22014 * (i - 128) + RGB_HALF;
kVToG[i] = -45773 * (i - 128);
kUToB[i] = (113618 * (i - 128) + RGB_HALF) >> RGB_FRAC;
}
for (i = RGB_RANGE_MIN; i < RGB_RANGE_MAX; ++i) {
const int j = ((i - 16) * 76283 + RGB_HALF) >> RGB_FRAC;
kClip[i - RGB_RANGE_MIN] = (j < 0) ? 0 : (j > 255) ? 255 : j;
}
init_done = 1;
}
static void ToRGB(int y, int u, int v, uint32* const dst) {
const int r_off = kVToR[v];
const int g_off = (kVToG[v] + kUToG[u]) >> RGB_FRAC;
const int b_off = kUToB[u];
const int r = kClip[y + r_off - RGB_RANGE_MIN];
const int g = kClip[y + g_off - RGB_RANGE_MIN];
const int b = kClip[y + b_off - RGB_RANGE_MIN];
*dst = (r << RED_SHIFT) | (g << GREEN_SHIFT) | (b << BLUE_SHIFT);
}
static inline uint32 get_le32(const uint8* const data) {
return data[0] | (data[1] << 8) | (data[2] << 16) | (data[3] << 24);
}
/* Returns the difference (in dB) between two images represented in YUV format
*
* Input:
* Y1/U1/V1: The Y/U/V data of the first image
* Y2/U2/V2: The Y/U/V data of the second image
*
* Returns the PSNR (http://en.wikipedia.org/wiki/Peak_signal-to-noise_ratio)
* value computed between the two images
*/
double GetPSNRYuv(const uint8* Y1,
const uint8* U1,
const uint8* V1,
const uint8* Y2,
const uint8* U2,
const uint8* V2,
int y_width,
int y_height) {
int x, y, row_idx;
const int uv_width = ((y_width + 1) >> 1);
const int uv_height = ((y_height + 1) >> 1);
double sse = 0., count = 0.;
for (y = 0; y < y_height; ++y) {
count += y_width;
row_idx = y * y_width;
for (x = 0; x < y_width; ++x) {
double diff = Y1[row_idx + x] - Y2[row_idx + x];
sse += diff * diff;
}
}
for (y = 0; y < uv_height; ++y) {
count += 2 * uv_width;
row_idx = y * uv_width;
for (x = 0; x < uv_width; ++x) {
const double diff_U = U1[row_idx + x] - U2[row_idx + x];
const double diff_V = V1[row_idx + x] - V2[row_idx + x];
sse += diff_U * diff_U + diff_V * diff_V;
}
}
return -4.3429448 * log(sse / (255. * 255. * count));
}
/* Returns the difference (in dB) between two images. One represented
* using Y,U,V vectors and the other is webp image data.
* Input:
* Y1/U1/V1: The Y/U/V data of the first image
* imgdata: data buffer containing webp image
* imgdata_size: size of the imgdata buffer
*
* Returns the PSNR value computed between the two images
*/
double WebPGetPSNR(const uint8* Y1,
const uint8* U1,
const uint8* V1,
uint8* imgdata,
int imgdata_size) {
uint8* Y2 = NULL;
uint8* U2 = NULL;
uint8* V2 = NULL;
int w = 0, h = 0;
double psnr = 0;
WebPDecode(imgdata,
imgdata_size,
&Y2,
&U2,
&V2,
&w,
&h);
psnr = GetPSNRYuv(Y1, U1, V1, Y2, U2, V2, w, h);
free(Y2);
return psnr;
}
/*---------------------------------------------------------------------*
* Reading WebP *
*---------------------------------------------------------------------*/
/* RIFF layout is:
* 0ffset tag
* 0...3 "RIFF" 4-byte tag
* 4...7 size of image data (including metadata) starting at offset 8
* 8...11 "WEBP" our form-type signature
* 12..15 "VP8 " 4-byte tags, describing the raw video format used
* 16..19 size of the raw WebP image data, starting at offset 20
* 20.... the WebP bytes
* There can be extra chunks after the "VP8 " chunk (ICMT, ICOP, ...)
* All 32-bits sizes are in little-endian order.
* Note: chunk data must be padded to multiple of 2 in size
*/
int SkipRiffHeader(const uint8** data_ptr, int *data_size_ptr) {
/* 20 bytes RIFF header 10 bytes VP8 header */
const int kHeaderSize = (20 + 10);
uint32 chunk_size = 0xffffffffu;
if (*data_size_ptr >= kHeaderSize && !memcmp(*data_ptr, "RIFF", 4)) {
if (memcmp(*data_ptr + 8, "WEBP", 4)) {
return 0; /* wrong image file signature */
} else {
const uint32 riff_size = get_le32(*data_ptr + 4);
if (memcmp(*data_ptr + 12, "VP8 ", 4)) {
return 0; /* invalid compression format */
}
chunk_size = get_le32(*data_ptr + 16);
if ((chunk_size > riff_size + 8) || (chunk_size & 1)) {
return 0; /* inconsistent size information. */
}
/* We have a RIFF container. Skip it. */
*data_ptr += 20;
*data_size_ptr -= 20;
}
}
return chunk_size;
}
/* Generate RGBA row from an YUV row (with width upsampling of chrome data)
* Input:
* 1, 2, 3. y_src, u_src, v_src - Pointers to input Y, U, V row data
* respectively. We reuse these variables, they iterate over all pixels in
* the row.
* 4. y_width: width of the Y image plane (aka image width)
* Output:
* 5. rgb_sat: pointer to the output rgb row. We reuse this variable, it
* iterates over all pixels in the row.
*/
static void YUV420toRGBLine(uint8* y_src,
uint8* u_src,
uint8* v_src,
int y_width,
uint32* rgb_dst) {
int x;
for (x = 0; x < (y_width >> 1); ++x) {
const int U = u_src[0];
const int V = v_src[0];
ToRGB(y_src[0], U, V, rgb_dst);
ToRGB(y_src[1], U, V, rgb_dst + 1);
++u_src;
++v_src;
y_src += 2;
rgb_dst += 2;
}
if (y_width & 1) { /* Rightmost pixel */
ToRGB(y_src[0], (*u_src), (*v_src), rgb_dst);
}
}
/* Converts from YUV (with color subsampling) such as produced by the WebPDecode
* routine into 32 bits per pixel RGBA data array. This data array can be
* directly used by the Leptonica Pix in-memory image format.
* Input:
* 1, 2, 3. Y, U, V: the input data buffers
* 4. pixwpl: the desired words per line corresponding to the supplied
* output pixdata.
* 5. width, height: the dimensions of the image whose data resides in Y,
* U, V.
* Output:
* 6. pixdata: the output data buffer. Caller should allocate
* height * pixwpl bytes of memory before calling this routine.
*/
void YUV420toRGBA(uint8* Y,
uint8* U,
uint8* V,
int words_per_line,
int width,
int height,
uint32* pixdata) {
int y_width = width;
int y_stride = y_width;
int uv_width = ((y_width + 1) >> 1);
int uv_stride = uv_width;
int y;
if (!init_done)
InitTables();
/* note that the U, V upsampling in height is happening here as the U, V
* buffers sent to successive odd-even pair of lines is same.
*/
for (y = 0; y < height; ++y) {
YUV420toRGBLine(Y + y * y_stride,
U + (y >> 1) * uv_stride,
V + (y >> 1) * uv_stride,
width,
pixdata + y * words_per_line);
}
}
void gd_YUV420toRGBA(uint8* Y,
uint8* U,
uint8* V,
gdImagePtr im) {
int width = im->sx;
int height = im->sy;
int y_width = width;
int y_stride = y_width;
int uv_width = ((y_width + 1) >> 1);
int uv_stride = uv_width;
int y;
/* output im must be truecolor */
if (!im->trueColor) {
return;
}
if (!init_done)
InitTables();
/* note that the U, V upsampling in height is happening here as the U, V
* buffers sent to successive odd-even pair of lines is same.
*/
for (y = 0; y < height; ++y) {
YUV420toRGBLine(Y + y * y_stride,
U + (y >> 1) * uv_stride,
V + (y >> 1) * uv_stride,
width,
im->tpixels[y]);
}
}
static WebPResult VPXDecode(const uint8* data,
int data_size,
uint8** p_Y,
uint8** p_U,
uint8** p_V,
int* p_width,
int* p_height) {
vpx_codec_ctx_t dec;
vp8_postproc_cfg_t ppcfg;
WebPResult result = webp_failure;
if (!data || data_size <= 10 || !p_Y || !p_U || !p_V
|| *p_Y != NULL || *p_U != NULL || *p_V != NULL) {
return webp_failure;
}
if (vpx_codec_dec_init(&dec,
&vpx_codec_vp8_dx_algo, NULL, 0) != VPX_CODEC_OK) {
return webp_failure;
}
ppcfg.post_proc_flag = VP8_NOFILTERING;
vpx_codec_control(&dec, VP8_SET_POSTPROC, &ppcfg);
if (vpx_codec_decode(&dec, data, data_size, NULL, 0) == VPX_CODEC_OK) {
vpx_codec_iter_t iter = NULL;
vpx_image_t* const img = vpx_codec_get_frame(&dec, &iter);
if (img) {
int y_width = img->d_w;
int y_height = img->d_h;
int y_stride = y_width;
int uv_width = (y_width + 1) >> 1;
int uv_stride = uv_width;
int uv_height = ((y_height + 1) >> 1);
int y;
*p_width = y_width;
*p_height = y_height;
if ((*p_Y = (uint8 *)(calloc(y_stride * y_height
+ 2 * uv_stride * uv_height,
sizeof(uint8)))) != NULL) {
*p_U = *p_Y + y_height * y_stride;
*p_V = *p_U + uv_height * uv_stride;
for (y = 0; y < y_height; ++y) {
memcpy(*p_Y + y * y_stride,
img->planes[0] + y * img->stride[0],
y_width);
}
for (y = 0; y < uv_height; ++y) {
memcpy(*p_U + y * uv_stride,
img->planes[1] + y * img->stride[1],
uv_width);
memcpy(*p_V + y * uv_stride,
img->planes[2] + y * img->stride[2],
uv_width);
}
result = webp_success;
}
}
}
vpx_codec_destroy(&dec);
return result;
}
WebPResult WebPDecode(const uint8* data,
int data_size,
uint8** p_Y,
uint8** p_U,
uint8** p_V,
int* p_width,
int* p_height) {
const uint32 chunk_size = SkipRiffHeader(&data, &data_size);
if (!chunk_size) {
return webp_failure; /* unsupported RIFF header */
}
return VPXDecode(data, data_size, p_Y, p_U, p_V, p_width, p_height);
}
/*---------------------------------------------------------------------*
* Writing WebP *
*---------------------------------------------------------------------*/
/* Takes a pair of RGBA row data as input and generates 2 rows of Y data and one
* row of subsampled U, V data as output
* Input:
* 1, 2. rgb_line1, rgb_line2 - input rgba rows
* 3. width - image width
* Outout:
* 4, 5, 6: Output Y, U, V row
*/
static void RGBALinepairToYUV420(uint32* rgb_line1,
uint32* rgb_line2,
int width,
uint8* Y_dst1,
uint8* Y_dst2,
uint8* u_dst,
uint8* v_dst) {
int x;
for (x = (width >> 1); x > 0; --x) {
const int sum_r =
GetRed(rgb_line1 + 0) + GetRed(rgb_line1 + 1) +
GetRed(rgb_line2 + 0) + GetRed(rgb_line2 + 1);
const int sum_g =
GetGreen(rgb_line1 + 0) + GetGreen(rgb_line1 + 1) +
GetGreen(rgb_line2 + 0) + GetGreen(rgb_line2 + 1);
const int sum_b =
GetBlue(rgb_line1 + 0) + GetBlue(rgb_line1 + 1) +
GetBlue(rgb_line2 + 0) + GetBlue(rgb_line2 + 1);
Y_dst1[0] = GetLumaYfromPtr(rgb_line1 + 0);
Y_dst1[1] = GetLumaYfromPtr(rgb_line1 + 1);
Y_dst2[0] = GetLumaYfromPtr(rgb_line2 + 0);
Y_dst2[1] = GetLumaYfromPtr(rgb_line2 + 1);
*u_dst++ = GetChromaU(sum_r, sum_g, sum_b);
*v_dst++ = GetChromaV(sum_r, sum_g, sum_b);
rgb_line1 += 2;
rgb_line2 += 2;
Y_dst1 += 2;
Y_dst2 += 2;
}
if (width & 1) { /* rightmost pixel. */
const int sum_r = GetRed(rgb_line1) + GetRed(rgb_line2);
const int sum_g = GetGreen(rgb_line1) + GetGreen(rgb_line2);
const int sum_b = GetBlue(rgb_line1) + GetBlue(rgb_line2);
Y_dst1[0] = GetLumaYfromPtr(rgb_line1);
Y_dst2[0] = GetLumaYfromPtr(rgb_line2);
*u_dst = GetChromaU(2 * sum_r, 2 * sum_g, 2 * sum_b);
*v_dst = GetChromaV(2 * sum_r, 2 * sum_g, 2 * sum_b);
}
}
/* Generates Y, U, V data (with color subsampling) from 32 bits
* per pixel RGBA data buffer. The resulting YUV data can be directly fed into
* the WebPEncode routine.
* Input:
* 1. pixdatainput rgba data buffer
* 2. words per line corresponding to pixdata
* 3, 4. image width and height respectively
* Output:
* 5, 6, 7. Output YUV data buffers
*/
void gd_RGBAToYUV420(gdImagePtr im2,
uint8* Y,
uint8* U,
uint8* V) {
int y_width = im2->sx;
int y_height = im2->sy;
int y_stride = y_width;
int uv_width = ((y_width + 1) >> 1);
int uv_stride = uv_width;
int y;
gdImagePtr im = NULL;
int free_im = 0;
if (!im2->trueColor) {
/* Todo: Replace the color/YUV functions with our own and simplify
that should boost the conversion a bit as well, not only for
palette image. */
im = gdImageCreateTrueColor(im2->sx, im2->sy);
if (!im) {
php_gd_error("gd-webp error: cannot convert palette input to truecolor");
return;
}
gdImageCopy(im, im2, 0, 0, 0, 0, im->sx, im->sy);
free_im = 1;
} else {
im = im2;
}
for (y = 0; y < (y_height >> 1); ++y) {
RGBALinepairToYUV420(im->tpixels[2 * y],
im->tpixels[2 * y + 1],
y_width,
Y + 2 * y * y_stride,
Y + (2 * y + 1) * y_stride,
U + y * uv_stride,
V + y * uv_stride);
}
if (y_height & 1) {
RGBALinepairToYUV420(im->tpixels[y_height - 1],
im->tpixels[y_height - 1],
y_width,
Y + (y_height - 1) * y_stride,
Y + (y_height - 1) * y_stride,
U + (y_height >> 1) * uv_stride,
V + (y_height >> 1) * uv_stride);
}
if (free_im) {
gdImageDestroy(im);
}
}
/* Generates Y, U, V data (with color subsampling) from 32 bits
* per pixel RGBA data buffer. The resulting YUV data can be directly fed into
* the WebPEncode routine.
* Input:
* 1. pixdatainput rgba data buffer
* 2. words per line corresponding to pixdata
* 3, 4. image width and height respectively
* Output:
* 5, 6, 7. Output YUV data buffers
*/
void RGBAToYUV420(uint32* pixdata,
int words_per_line,
int width,
int height,
uint8* Y,
uint8* U,
uint8* V) {
int y_width = width;
int y_height = height;
int y_stride = y_width;
int uv_width = ((y_width + 1) >> 1);
int uv_stride = uv_width;
int y;
for (y = 0; y < (y_height >> 1); ++y) {
RGBALinepairToYUV420(pixdata + 2 * y * words_per_line,
pixdata + (2 * y + 1) * words_per_line,
y_width,
Y + 2 * y * y_stride,
Y + (2 * y + 1) * y_stride,
U + y * uv_stride,
V + y * uv_stride);
}
if (y_height & 1) {
RGBALinepairToYUV420(pixdata + (y_height - 1) * words_per_line,
pixdata + (y_height - 1) * words_per_line,
y_width,
Y + (y_height - 1) * y_stride,
Y + (y_height - 1) * y_stride,
U + (y_height >> 1) * uv_stride,
V + (y_height >> 1) * uv_stride);
}
}
static int codec_ctl(vpx_codec_ctx_t *enc,
enum vp8e_enc_control_id id,
int value) {
const vpx_codec_err_t res = vpx_codec_control_(enc, id, value);
if (res != VPX_CODEC_OK) {
return webp_failure;
}
return webp_success;
}
static void SetupParams(vpx_codec_enc_cfg_t* cfg,
int QP) {
cfg->g_threads = 2;
cfg->rc_min_quantizer = QP;
cfg->rc_max_quantizer = QP;
cfg->kf_mode = VPX_KF_FIXED;
}
/* VPXEncode: Takes a Y, U, V data buffers (with color components U and V
* subsampled to 1/2 resolution) and generates the VPX string.
* Output VPX string is placed in the *p_out buffer. container_size
* indicates number of bytes to be left blank at the beginning of
* *p_out buffer to accommodate for a container header.
*
* Return: success/failure
*/
static WebPResult VPXEncode(const uint8* Y,
const uint8* U,
const uint8* V,
int y_width,
int y_height,
int y_stride,
int uv_width,
int uv_height,
int uv_stride,
int QP,
int container_size,
unsigned char** p_out,
int* p_out_size_bytes) {
vpx_codec_iface_t* iface = &vpx_codec_vp8_cx_algo;
vpx_codec_err_t res;
vpx_codec_enc_cfg_t cfg;
vpx_codec_ctx_t enc;
WebPResult result = webp_failure;
vpx_image_t img;
*p_out = NULL;
*p_out_size_bytes = 0;
/* validate input parameters. */
if (!p_out || !Y || !U || !V
|| y_width <= 0 || y_height <= 0 || uv_width <= 0 || uv_height <= 0
|| y_stride < y_width || uv_stride < uv_width
|| QP < 0 || QP > 63) {
return webp_failure;
}
res = vpx_codec_enc_config_default(iface, &cfg, 0);
if (res != VPX_CODEC_OK) {
return webp_failure;
}
SetupParams(&cfg, QP);
cfg.g_w = y_width;
cfg.g_h = y_height;
res = vpx_codec_enc_init(&enc, iface, &cfg, 0);
if (res == VPX_CODEC_OK) {
codec_ctl(&enc, VP8E_SET_CPUUSED, 3);
codec_ctl(&enc, VP8E_SET_NOISE_SENSITIVITY, 0);
codec_ctl(&enc, VP8E_SET_SHARPNESS, 0);
codec_ctl(&enc, VP8E_SET_ENABLEAUTOALTREF, 0);
codec_ctl(&enc, VP8E_SET_ARNR_MAXFRAMES, 0);
codec_ctl(&enc, VP8E_SET_ARNR_TYPE, 0);
codec_ctl(&enc, VP8E_SET_ARNR_STRENGTH, 0);
codec_ctl(&enc, VP8E_SET_STATIC_THRESHOLD, 0);
codec_ctl(&enc, VP8E_SET_TOKEN_PARTITIONS, 2);
vpx_img_wrap(&img, VPX_IMG_FMT_I420,
y_width, y_height, 16, (uint8*)(Y));
img.planes[VPX_PLANE_Y] = (uint8*)(Y);
img.planes[VPX_PLANE_U] = (uint8*)(U);
img.planes[VPX_PLANE_V] = (uint8*)(V);
img.stride[VPX_PLANE_Y] = y_stride;
img.stride[VPX_PLANE_U] = uv_stride;
img.stride[VPX_PLANE_V] = uv_stride;
res = vpx_codec_encode(&enc, &img, 0, 1, 0, VPX_DL_BEST_QUALITY);
if (res == VPX_CODEC_OK) {
vpx_codec_iter_t iter = NULL;
const vpx_codec_cx_pkt_t* pkt = vpx_codec_get_cx_data(&enc, &iter);
if (pkt != NULL) {
*p_out = (unsigned char*)(calloc(container_size + pkt->data.frame.sz,
1));
memcpy(*p_out + container_size,
(const void*)(pkt->data.frame.buf),
pkt->data.frame.sz);
*p_out_size_bytes = container_size + pkt->data.frame.sz;
result = webp_success;
}
}
}
vpx_codec_destroy(&enc);
return result;
}
WebPResult WebPEncode(const uint8* Y,
const uint8* U,
const uint8* V,
int y_width,
int y_height,
int y_stride,
int uv_width,
int uv_height,
int uv_stride,
int QP,
unsigned char** p_out,
int* p_out_size_bytes,
double *psnr) {
const int kRiffHeaderSize = 20;
if (VPXEncode(Y, U, V,
y_width, y_height, y_stride,
uv_width, uv_height, uv_stride,
QP, kRiffHeaderSize,
p_out, p_out_size_bytes) != webp_success) {
return webp_failure;
} else {
/* Write RIFF header */
const int img_size_bytes = *p_out_size_bytes - kRiffHeaderSize;
const int chunk_size = (img_size_bytes + 1) & ~1; /* make size even */
const int riff_size = chunk_size + 12;
const uint8_t kRiffHeader[20] = { 'R', 'I', 'F', 'F',
(riff_size >> 0) & 255,
(riff_size >> 8) & 255,
(riff_size >> 16) & 255,
(riff_size >> 24) & 255,
'W', 'E', 'B', 'P',
'V', 'P', '8', ' ',
(chunk_size >> 0) & 255,
(chunk_size >> 8) & 255,
(chunk_size >> 16) & 255,
(chunk_size >> 24) & 255 };
memcpy(*p_out, kRiffHeader, kRiffHeaderSize);
if (img_size_bytes & 1) { /* write a padding byte */
const int new_size = *p_out_size_bytes + 1;
unsigned char* p = (unsigned char*)realloc(*p_out, new_size);
if (p == NULL) {
free(*p_out);
*p_out = NULL;
*p_out_size_bytes = 0;
return webp_failure;
}
p[new_size - 1] = 0;
*p_out = p;
*p_out_size_bytes = new_size;
}
if (psnr) {
*psnr = WebPGetPSNR(Y, U, V, *p_out, *p_out_size_bytes);
}
return webp_success;
}
}
void AdjustColorspace(uint8* Y, uint8* U, uint8* V, int width, int height) {
int y_width = width;
int y_height = height;
int y_stride = y_width;
int uv_width = ((y_width + 1) >> 1);
int uv_height = ((y_height + 1) >> 1);
int uv_stride = uv_width;
int x, y;
/* convert luma */
for (y = 0; y < y_height; ++y) {
uint8* const Yrow = Y + y * y_stride;
for (x = 0; x < y_width; ++x) {
/* maps [0..255] to [16..235] */
Yrow[x] = ((Yrow[x] * 55 + 32) >> 6) + 16;
}
}
/* convert chroma */
for (y = 0; y < uv_height; ++y) {
uint8* const Urow = U + y * uv_stride;
uint8* const Vrow = V + y * uv_stride;
for (x = 0; x < uv_width; ++x) {
/* maps [0..255] to [16..240] */
Urow[x] = (((Urow[x] - 127) * 7) >> 3) + 128;
Vrow[x] = (((Vrow[x] - 127) * 7) >> 3) + 128;
}
}
}
void AdjustColorspaceBack(uint8* Y, uint8* U, uint8* V, int width, int height) {
int y_width = width;
int y_height = height;
int y_stride = y_width;
int uv_width = ((y_width + 1) >> 1);
int uv_height = ((y_height + 1) >> 1);
int uv_stride = uv_width;
int x, y;
/* convert luma */
for (y = 0; y < y_height; ++y) {
uint8* const Yrow = Y + y * y_stride;
for (x = 0; x < y_width; ++x) {
/* maps [16..235] to [0..255] */
const int v = ((Yrow[x] - 16) * 149 + 64) >> 7;
Yrow[x] = (v < 0) ? 0 : (v > 255) ? 255u : v;
}
}
/* convert chroma */
for (y = 0; y < uv_height; ++y) {
uint8* const Urow = U + y * uv_stride;
uint8* const Vrow = V + y * uv_stride;
for (x = 0; x < uv_width; ++x) {
/* maps [0..255] to [16..240] */
const int ru = (((Urow[x] - 128) * 73) >> 6) + 128;
const int rv = (((Vrow[x] - 128) * 73) >> 6) + 128;
Urow[x] = (ru < 0) ? 0 : (ru > 255) ? 255u : ru;
Vrow[x] = (rv < 0) ? 0 : (rv > 255) ? 255u : rv;
}
}
}
WebPResult WebPGetInfo(const uint8* data,
int data_size,
int *width,
int *height) {
const uint32 chunk_size = SkipRiffHeader(&data, &data_size);
if (width) *width = 0;
if (height) *height = 0;
if (!chunk_size) {
return webp_failure; /* unsupported RIFF header */
}
/* Validate raw video data */
if (data_size < 10) {
return webp_failure; /* not enough data */
}
/* check signature */
if (data[3] != 0x9d || data[4] != 0x01 || data[5] != 0x2a) {
return webp_failure; /* Wrong signature. */
} else {
const uint32 bits = data[0] | (data[1] << 8) | (data[2] << 16);
if ((bits & 1)) { /* Not a keyframe. */
return webp_failure;
} else {
const int profile = (bits >> 1) & 7;
const int show_frame = (bits >> 4) & 1;
const uint32 partition_length = (bits >> 5);
if (profile > 3) {
return webp_failure; /* unknown profile */
}
if (!show_frame) {
return webp_failure; /* first frame is invisible! */
}
if (partition_length >= chunk_size) {
return webp_failure; /* inconsistent size information. */
} else {
const int w = ((data[7] << 8) | data[6]) & 0x3fff;
const int h = ((data[9] << 8) | data[8]) & 0x3fff;
if (width) *width = w;
if (height) *height = h;
return webp_success;
}
}
}
return webp_failure;
}
#endif /* HAVE_LIBVPX */

View File

@ -1,181 +0,0 @@
/*===========================================================================*
- Copyright 2010 Google Inc.
-
- This code is licensed under the same terms as WebM:
- Software License Agreement: http://www.webmproject.org/license/software/
- Additional IP Rights Grant: http://www.webmproject.org/license/additional/
*===========================================================================*/
/*
* Encoding/Decoding of WebP still image compression format.
*
* 1. WebPDecode: Takes an array of bytes (string) corresponding to the WebP
* encoded image and generates output in the YUV format with
* the color components U, V subsampled to 1/2 resolution along
* each dimension.
*
* 2. YUV420toRGBA: Converts from YUV (with color subsampling) such as produced
* by the WebPDecode routine into 32 bits per pixel RGBA data
* array. This data array can be directly used by the Leptonica
* Pix in-memory image format.
*
* 3. WebPEncode: Takes a Y, U, V data buffers (with color components U and V
* subsampled to 1/2 resolution) and generates the WebP string
*
* 4. RGBAToYUV420: Generates Y, U, V data (with color subsampling) from 32 bits
* per pixel RGBA data buffer. The resulting YUV data can be
* directly fed into the WebPEncode routine.
*
* 5. AdjustColorspace:
*
* 6. AdjustColorspaceBack:
*/
#ifndef THIRD_PARTY_VP8_VP8IMG_H_
#define THIRD_PARTY_VP8_VP8IMG_H_
#ifdef __cplusplus
extern "C" {
#endif /* __cplusplus */
typedef unsigned char uint8;
typedef unsigned int uint32;
typedef enum WebPResultType {
webp_success = 0,
webp_failure = -1
} WebPResult;
/* Takes an array of bytes (string) corresponding to the WebP
* encoded image and generates output in the YUV format with
* the color components U, V subsampled to 1/2 resolution along
* each dimension.
* Input:
* 1. data: the WebP data stream (array of bytes)
* 2. data_size: count of bytes in the WebP data stream
*
* Output:
* 3. p_Y/p_U/p_V : pointer to the Y/U/V data buffer (this routine will
* allocate memory for the buffer, fill the buffer with
* appropriate data and transfer owner ship of the buffer
* to caller. Caller is responsible for freeing the memory).
* Note that the memory for Y, U, V buffers is alloacted
* in one chunk, hence one should call free(*p_Y) only.
* Do not try to free the U and V buffers.
*
* 6. p_width: this routine returns the width of the decoded image here
* 7. p_height: this routine returns the width of the decoded image here
* Return: success/failure
*/
WebPResult WebPDecode(const uint8* data,
int data_size,
uint8** p_Y,
uint8** p_U,
uint8** p_V,
int* p_width,
int* p_height);
/* WebPEncode: Takes a Y, U, V data buffers (with color components U and V
* subsampled to 1/2 resolution) and generates the WebP string.
* Input:
* 1, 2, 3. Y, U, V: The input YUV data buffers
* 4, 5. y_width, y_height: The width and height of the image whose data
* is in Y, U, V. This matches the Y plane. The U
* and V planes typically have 1/2 width and
* height.
* 6. y_stride: The width (in bytes) of one row of Y data. This may not
* match width if there is end of row padding (e.g., for 32
* bit row alignment).
* 7. QP: the quantization parameter. This parameter controls the
* compression vs quality tradeoff. Use smaller numbers for better
* quality (compression will be lesser) and vice versa. 20 is a
* good optimal value.
* Output:
* 8. p_out: the output array of bytes corresponding to the encoded WebP
* image. This routine allocates memory for the buffer, fills it
* with appropriate values and transfers ownership to caller.
* Caller responsible for freeing of memory.
* Return: success/failure
*/
WebPResult WebPEncode(const uint8* Y,
const uint8* U,
const uint8* V,
int y_width,
int y_height,
int y_stride,
int uv_width,
int uv_height,
int uv_stride,
int QP,
unsigned char** p_out,
int* p_out_size_bytes,
double* psnr);
/* Converts from YUV (with color subsampling) such as produced by the WebPDecode
* routine into 32 bits per pixel RGBA data array. This data array can be
* directly used by the Leptonica Pix in-memory image format.
* Input:
* 1, 2, 3. Y, U, V: the input data buffers
* 4. pixwpl: the desired words per line corresponding to the supplied
* output pixdata.
* 5. width, height: the dimensions of the image whose data resides in Y,
* U, V.
* Output:
* 6. pixdata: the output data buffer. Caller should allocate
* height * pixwpl bytes of memory before calling this routine.
*/
void YUV420toRGBA(uint8* Y,
uint8* U,
uint8* V,
int words_per_line,
int width,
int height,
uint32* pixdata);
/* Generates Y, U, V data (with color subsampling) from 32 bits
* per pixel RGBA data buffer. The resulting YUV data can be directly fed into
* the WebPEncode routine.
* Input:
* 1. pix data input rgba data buffer
* 2. words per line corresponding to pixdata
* 3, 4. image width and height respectively
* Output:
* 5, 6, 7. Output YUV data buffers
*/
void RGBAToYUV420(uint32* pixdata,
int words_per_line,
int width,
int height,
uint8* Y,
uint8* U,
uint8* V);
/* This function adjust from YUV420J (jpeg decoding) to YUV420 (webp input)
* Hints: http://en.wikipedia.org/wiki/YCbCr
*/
void AdjustColorspace(uint8* Y, uint8* U, uint8* V, int width, int height);
/* Inverse function: convert from YUV420 to YUV420J */
void AdjustColorspaceBack(uint8* Y, uint8* U, uint8* V, int width, int height);
/* Checks WebP image header and outputs height and width information of
* the image
*
* Input:
* 1. data: the WebP data stream (array of bytes)
* 2. data_size: count of bytes in the WebP data stream
*
* Outut:
* width/height: width and height of the image
*
* Return: success/failure
*/
WebPResult WebPGetInfo(const uint8* data,
int data_size,
int *width,
int *height);
#ifdef __cplusplus
}
#endif /* __cplusplus */
#endif /* THIRD_PARTY_VP8_VP8IMG_H_ */