--- image.c 2003/05/17 13:36:18 1.26.2.4 +++ image.c 2004/01/30 18:53:50 1.26.2.14 @@ -1,68 +1,42 @@ /************************************************************************** * - * XVID MPEG-4 VIDEO CODEC - * image stuff + * XVID MPEG-4 VIDEO CODEC + * - Image management functions - * - * This program is an implementation of a part of one or more MPEG-4 - * Video tools as specified in ISO/IEC 14496-2 standard. Those intending - * to use this software module in hardware or software products are - * advised that its use may infringe existing patents or copyrights, and - * any such use would be at such party's own risk. The original - * developer of this software module and his/her company, and subsequent - * editors and their companies, will have no liability for use of this - * software or modifications or derivatives thereof. + * Copyright(C) 2001-2003 Peter Ross * - * This program is free software; you can redistribute it and/or modify - * it under the terms of the GNU General Public License as published by - * the Free Software Foundation; either version 2 of the License, or - * (at your option) any later version. + * This program is free software ; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation ; either version 2 of the License, or + * (at your option) any later version. * - * This program is distributed in the hope that it will be useful, - * but WITHOUT ANY WARRANTY; without even the implied warranty of - * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the - * GNU General Public License for more details. + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY ; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. * - * You should have received a copy of the GNU General Public License - * along with this program; if not, write to the Free Software - * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. + * You should have received a copy of the GNU General Public License + * along with this program ; if not, write to the Free Software + * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA * - *************************************************************************/ - -/************************************************************************** - * - * History: + * $Id: image.c,v 1.26.2.14 2004/01/30 18:53:50 chl Exp $ * - * 05.10.2002 support for interpolated images in qpel mode - Isibaar - * 01.05.2002 BFRAME image-based u,v interpolation - * 22.04.2002 added some B-frame support - * 14.04.2002 added image_dump_yuvpgm(), added image_mad() - * XVID_CSP_USER input support - * 09.04.2002 PSNR calculations - Isibaar - * 06.04.2002 removed interlaced edging from U,V blocks (as per spec) - * 26.03.2002 interlacing support (field-based edging in set_edges) - * 26.01.2002 rgb555, rgb565 - * 07.01.2001 commented u,v interpolation (not required for uv-block-based) - * 23.12.2001 removed #ifdefs, added function pointers + init_common() - * 22.12.2001 cpu #ifdefs - * 19.12.2001 image_dump(); useful for debugging - * 6.12.2001 inital version; (c)2001 peter ross - * - *************************************************************************/ + ****************************************************************************/ #include -#include // memcpy, memset +#include /* memcpy, memset */ #include #include "../portab.h" -#include "../global.h" // XVID_CSP_XXX's -#include "../xvid.h" // XVID_CSP_XXX's +#include "../global.h" /* XVID_CSP_XXX's */ +#include "../xvid.h" /* XVID_CSP_XXX's */ #include "image.h" #include "colorspace.h" #include "interpolate8x8.h" #include "reduced.h" #include "../utils/mem_align.h" -#include "font.h" // XXX: remove later +#include "font.h" /* XXX: remove later */ #define SAFETY 64 #define EDGE_SIZE2 (EDGE_SIZE/2) @@ -75,29 +49,31 @@ { const uint32_t edged_width2 = edged_width / 2; const uint32_t edged_height2 = edged_height / 2; - uint32_t i; image->y = xvid_malloc(edged_width * (edged_height + 1) + SAFETY, CACHE_LINE); if (image->y == NULL) { return -1; } - - for (i = 0; i < edged_width * edged_height + SAFETY; i++) { - image->y[i] = 0; - } + memset(image->y, 0, edged_width * (edged_height + 1) + SAFETY); image->u = xvid_malloc(edged_width2 * edged_height2 + SAFETY, CACHE_LINE); if (image->u == NULL) { xvid_free(image->y); + image->y = NULL; return -1; } + memset(image->u, 0, edged_width2 * edged_height2 + SAFETY); + image->v = xvid_malloc(edged_width2 * edged_height2 + SAFETY, CACHE_LINE); if (image->v == NULL) { xvid_free(image->u); + image->u = NULL; xvid_free(image->y); + image->y = NULL; return -1; } + memset(image->v, 0, edged_width2 * edged_height2 + SAFETY); image->y += EDGE_SIZE * edged_width + EDGE_SIZE; image->u += EDGE_SIZE2 * edged_width2 + EDGE_SIZE2; @@ -117,12 +93,15 @@ if (image->y) { xvid_free(image->y - (EDGE_SIZE * edged_width + EDGE_SIZE)); + image->y = NULL; } if (image->u) { xvid_free(image->u - (EDGE_SIZE2 * edged_width2 + EDGE_SIZE2)); + image->u = NULL; } if (image->v) { xvid_free(image->v - (EDGE_SIZE2 * edged_width2 + EDGE_SIZE2)); + image->v = NULL; } } @@ -157,7 +136,7 @@ uint32_t height) { const uint32_t edged_width2 = edged_width / 2; - const uint32_t width2 = width / 2; + uint32_t width2; uint32_t i; uint8_t *dst; uint8_t *src; @@ -166,6 +145,12 @@ dst = image->y - (EDGE_SIZE + EDGE_SIZE * edged_width); src = image->y; + /* According to the Standard Clause 7.6.4, padding is done starting at 16 + * pixel width and height multiples */ + width = (width+15)&~15; + height = (height+15)&~15; + width2 = width/2; + for (i = 0; i < EDGE_SIZE; i++) { memset(dst, *src, EDGE_SIZE); memcpy(dst + EDGE_SIZE, src, width); @@ -191,7 +176,7 @@ } -//U + /* U */ dst = image->u - (EDGE_SIZE2 + EDGE_SIZE2 * edged_width2); src = image->u; @@ -219,7 +204,7 @@ } -// V + /* V */ dst = image->v - (EDGE_SIZE2 + EDGE_SIZE2 * edged_width2); src = image->v; @@ -247,7 +232,7 @@ } } -// bframe encoding requires image-based u,v interpolation +/* bframe encoding requires image-based u,v interpolation */ void image_interpolate(const IMAGE * refn, IMAGE * refh, @@ -258,16 +243,14 @@ uint32_t quarterpel, uint32_t rounding) { - const uint32_t offset = EDGE_SIZE2 * (edged_width + 1); // we only interpolate half of the edge area + const uint32_t offset = EDGE_SIZE2 * (edged_width + 1); /* we only interpolate half of the edge area */ const uint32_t stride_add = 7 * edged_width; -/* -#ifdef BFRAMES +#if 0 const uint32_t edged_width2 = edged_width / 2; const uint32_t edged_height2 = edged_height / 2; const uint32_t offset2 = EDGE_SIZE2 * (edged_width2 + 1); const uint32_t stride_add2 = 7 * edged_width2; #endif -*/ uint8_t *n_ptr, *h_ptr, *v_ptr, *hv_ptr; uint32_t x, y; @@ -275,15 +258,15 @@ n_ptr = refn->y; h_ptr = refh->y; v_ptr = refv->y; - hv_ptr = refhv->y; n_ptr -= offset; h_ptr -= offset; v_ptr -= offset; - hv_ptr -= offset; + /* Note we initialize the hv pointer later, as we can optimize code a bit + * doing it down to up in quarterpel and up to down in halfpel */ if(quarterpel) { - + for (y = 0; y < (edged_height - EDGE_SIZE); y += 8) { for (x = 0; x < (edged_width - EDGE_SIZE); x += 8) { interpolate8x8_6tap_lowpass_h(h_ptr, n_ptr, edged_width, rounding); @@ -293,7 +276,7 @@ h_ptr += 8; v_ptr += 8; } - + n_ptr += EDGE_SIZE; h_ptr += EDGE_SIZE; v_ptr += EDGE_SIZE; @@ -303,24 +286,25 @@ n_ptr += stride_add; } - h_ptr = refh->y; - h_ptr -= offset; + h_ptr = refh->y + (edged_height - EDGE_SIZE - EDGE_SIZE2)*edged_width - EDGE_SIZE2; + hv_ptr = refhv->y + (edged_height - EDGE_SIZE - EDGE_SIZE2)*edged_width - EDGE_SIZE2; for (y = 0; y < (edged_height - EDGE_SIZE); y = y + 8) { + hv_ptr -= stride_add; + h_ptr -= stride_add; + hv_ptr -= EDGE_SIZE; + h_ptr -= EDGE_SIZE; + for (x = 0; x < (edged_width - EDGE_SIZE); x = x + 8) { + hv_ptr -= 8; + h_ptr -= 8; interpolate8x8_6tap_lowpass_v(hv_ptr, h_ptr, edged_width, rounding); - hv_ptr += 8; - h_ptr += 8; } - - hv_ptr += EDGE_SIZE; - h_ptr += EDGE_SIZE; - - hv_ptr += stride_add; - h_ptr += stride_add; } - } - else { + } else { + + hv_ptr = refhv->y; + hv_ptr -= offset; for (y = 0; y < (edged_height - EDGE_SIZE); y += 8) { for (x = 0; x < (edged_width - EDGE_SIZE); x += 8) { @@ -333,7 +317,7 @@ v_ptr += 8; hv_ptr += 8; } - + h_ptr += EDGE_SIZE; v_ptr += EDGE_SIZE; hv_ptr += EDGE_SIZE; @@ -405,13 +389,13 @@ /* interpolate_halfpel_h( refh->y - offset, - refn->y - offset, + refn->y - offset, edged_width, edged_height, rounding); interpolate_halfpel_v( refv->y - offset, - refn->y - offset, + refn->y - offset, edged_width, edged_height, rounding); @@ -427,38 +411,38 @@ interpolate_halfpel_h( refh->u - offset, - refn->u - offset, + refn->u - offset, edged_width / 2, edged_height / 2, rounding); interpolate_halfpel_v( refv->u - offset, - refn->u - offset, + refn->u - offset, edged_width / 2, edged_height / 2, rounding); interpolate_halfpel_hv( refhv->u - offset, - refn->u - offset, + refn->u - offset, edged_width / 2, edged_height / 2, rounding); interpolate_halfpel_h( refh->v - offset, - refn->v - offset, + refn->v - offset, edged_width / 2, edged_height / 2, rounding); interpolate_halfpel_v( refv->v - offset, - refn->v - offset, + refn->v - offset, edged_width / 2, edged_height / 2, rounding); interpolate_halfpel_hv( refhv->v - offset, - refn->v - offset, + refn->v - offset, edged_width / 2, edged_height / 2, rounding); */ @@ -485,9 +469,9 @@ #define IMG_U(Y,X) img->u[(Y)*edged_width/2 + (X)] #define IMG_V(Y,X) img->v[(Y)*edged_width/2 + (X)] - if (IS_PURE(IMG_Y(y*2 ,x*2 )) && + if (IS_PURE(IMG_Y(y*2 ,x*2 )) && IS_PURE(IMG_Y(y*2 ,x*2+1)) && - IS_PURE(IMG_Y(y*2+1,x*2 )) && + IS_PURE(IMG_Y(y*2+1,x*2 )) && IS_PURE(IMG_Y(y*2+1,x*2+1))) { IMG_U(y,x) = (IMG_U(y,x-1) + IMG_U(y-1, x) + IMG_U(y, x+1) + IMG_U(y+1, x)) / 4; @@ -500,8 +484,8 @@ #undef IMG_U #undef IMG_V } - - DPRINTF(XVID_DEBUG_DEBUG,"chroma_optimized_pixels = %i/%i", pixels, width*height/4); + + DPRINTF(XVID_DEBUG_DEBUG,"chroma_optimized_pixels = %i/%i\n", pixels, width*height/4); } @@ -514,7 +498,7 @@ and two unoptimized/plain-c areas (pixel width divisible by 2) */ -static void +static void safe_packed_conv(uint8_t * x_ptr, int x_stride, uint8_t * y_ptr, uint8_t * u_ptr, uint8_t * v_ptr, int y_stride, int uv_stride, @@ -561,12 +545,14 @@ const int edged_width2 = edged_width/2; const int width2 = width/2; const int height2 = height/2; - //const int height_signed = (csp & XVID_CSP_VFLIP) ? -height : height; +#if 0 + const int height_signed = (csp & XVID_CSP_VFLIP) ? -height : height; +#endif switch (csp & ~XVID_CSP_VFLIP) { case XVID_CSP_RGB555: safe_packed_conv( - src[0], src_stride[0], image->y, image->u, image->v, + src[0], src_stride[0], image->y, image->u, image->v, edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), interlacing?rgb555i_to_yv12 :rgb555_to_yv12, interlacing?rgb555i_to_yv12_c:rgb555_to_yv12_c, 2); @@ -574,7 +560,7 @@ case XVID_CSP_RGB565: safe_packed_conv( - src[0], src_stride[0], image->y, image->u, image->v, + src[0], src_stride[0], image->y, image->u, image->v, edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), interlacing?rgb565i_to_yv12 :rgb565_to_yv12, interlacing?rgb565i_to_yv12_c:rgb565_to_yv12_c, 2); @@ -583,7 +569,7 @@ case XVID_CSP_BGR: safe_packed_conv( - src[0], src_stride[0], image->y, image->u, image->v, + src[0], src_stride[0], image->y, image->u, image->v, edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), interlacing?bgri_to_yv12 :bgr_to_yv12, interlacing?bgri_to_yv12_c:bgr_to_yv12_c, 3); @@ -591,7 +577,7 @@ case XVID_CSP_BGRA: safe_packed_conv( - src[0], src_stride[0], image->y, image->u, image->v, + src[0], src_stride[0], image->y, image->u, image->v, edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), interlacing?bgrai_to_yv12 :bgra_to_yv12, interlacing?bgrai_to_yv12_c:bgra_to_yv12_c, 4); @@ -599,7 +585,7 @@ case XVID_CSP_ABGR : safe_packed_conv( - src[0], src_stride[0], image->y, image->u, image->v, + src[0], src_stride[0], image->y, image->u, image->v, edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), interlacing?abgri_to_yv12 :abgr_to_yv12, interlacing?abgri_to_yv12_c:abgr_to_yv12_c, 4); @@ -607,15 +593,23 @@ case XVID_CSP_RGBA : safe_packed_conv( - src[0], src_stride[0], image->y, image->u, image->v, + src[0], src_stride[0], image->y, image->u, image->v, edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), interlacing?rgbai_to_yv12 :rgba_to_yv12, interlacing?rgbai_to_yv12_c:rgba_to_yv12_c, 4); break; + + case XVID_CSP_ARGB: + safe_packed_conv( + src[0], src_stride[0], image->y, image->u, image->v, + edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), + interlacing?argbi_to_yv12 : argb_to_yv12, + interlacing?argbi_to_yv12_c: argb_to_yv12_c, 4); + break; case XVID_CSP_YUY2: safe_packed_conv( - src[0], src_stride[0], image->y, image->u, image->v, + src[0], src_stride[0], image->y, image->u, image->v, edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), interlacing?yuyvi_to_yv12 :yuyv_to_yv12, interlacing?yuyvi_to_yv12_c:yuyv_to_yv12_c, 2); @@ -623,7 +617,7 @@ case XVID_CSP_YVYU: /* u/v swapped */ safe_packed_conv( - src[0], src_stride[0], image->y, image->v, image->y, + src[0], src_stride[0], image->y, image->v, image->y, edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), interlacing?yuyvi_to_yv12 :yuyv_to_yv12, interlacing?yuyvi_to_yv12_c:yuyv_to_yv12_c, 2); @@ -631,28 +625,27 @@ case XVID_CSP_UYVY: safe_packed_conv( - src[0], src_stride[0], image->y, image->u, image->v, + src[0], src_stride[0], image->y, image->u, image->v, edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), interlacing?uyvyi_to_yv12 :uyvy_to_yv12, interlacing?uyvyi_to_yv12_c:uyvy_to_yv12_c, 2); break; - case XVID_CSP_I420: + case XVID_CSP_YV12: /* YCrCb == internal colorspace for MPEG */ yv12_to_yv12(image->y, image->u, image->v, edged_width, edged_width2, src[0], src[0] + src_stride[0]*height, src[0] + src_stride[0]*height + (src_stride[0]/2)*height2, src_stride[0], src_stride[0]/2, width, height, (csp & XVID_CSP_VFLIP)); - break - ; - case XVID_CSP_YV12: /* u/v swapped */ + break; + + case XVID_CSP_I420: /* YCbCr == U and V plane swapped */ yv12_to_yv12(image->y, image->v, image->u, edged_width, edged_width2, src[0], src[0] + src_stride[0]*height, src[0] + src_stride[0]*height + (src_stride[0]/2)*height2, src_stride[0], src_stride[0]/2, width, height, (csp & XVID_CSP_VFLIP)); break; - case XVID_CSP_USER: - /*XXX: support for different u & v strides */ + case XVID_CSP_PLANAR: /* YCbCr with arbitrary pointers and different strides for Y and UV */ yv12_to_yv12(image->y, image->u, image->v, edged_width, edged_width2, - src[0], src[1], src[2], src_stride[0], src_stride[1], + src[0], src[1], src[2], src_stride[0], src_stride[1], /* v: dst_stride[2] not yet supported */ width, height, (csp & XVID_CSP_VFLIP)); break; @@ -672,21 +665,21 @@ int pad_width = 16 - (width&15); for (i = 0; i < height; i++) { - memset(image->y + i*edged_width + width, + memset(image->y + i*edged_width + width, *(image->y + i*edged_width + width - 1), pad_width); } for (i = 0; i < height/2; i++) { - memset(image->u + i*edged_width2 + width2, + memset(image->u + i*edged_width2 + width2, *(image->u + i*edged_width2 + width2 - 1),pad_width/2); - memset(image->v + i*edged_width2 + width2, + memset(image->v + i*edged_width2 + width2, *(image->v + i*edged_width2 + width2 - 1),pad_width/2); } } if (height & 15) { - int pad_height = 16 - (height&15); + int pad_height = 16 - (height&15); int length = ((width+15)/16)*16; int i; for (i = 0; i < pad_height; i++) @@ -782,6 +775,14 @@ interlacing?yv12_to_rgbai_c:yv12_to_rgba_c, 4); return 0; + case XVID_CSP_ARGB: + safe_packed_conv( + dst[0], dst_stride[0], image->y, image->u, image->v, + edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), + interlacing?yv12_to_argbi :yv12_to_argb, + interlacing?yv12_to_argbi_c:yv12_to_argb_c, 4); + return 0; + case XVID_CSP_YUY2: safe_packed_conv( dst[0], dst_stride[0], image->y, image->u, image->v, @@ -790,7 +791,7 @@ interlacing?yv12_to_yuyvi_c:yv12_to_yuyv_c, 2); return 0; - case XVID_CSP_YVYU: // u,v swapped + case XVID_CSP_YVYU: /* u,v swapped */ safe_packed_conv( dst[0], dst_stride[0], image->y, image->v, image->u, edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP), @@ -806,24 +807,24 @@ interlacing?yv12_to_uyvyi_c:yv12_to_uyvy_c, 2); return 0; - case XVID_CSP_I420: + case XVID_CSP_YV12: /* YCbCr == internal colorspace for MPEG */ yv12_to_yv12(dst[0], dst[0] + dst_stride[0]*height, dst[0] + dst_stride[0]*height + (dst_stride[0]/2)*height2, dst_stride[0], dst_stride[0]/2, image->y, image->u, image->v, edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP)); return 0; - case XVID_CSP_YV12: // u,v swapped + case XVID_CSP_I420: /* YCrCb == U and V plane swapped */ yv12_to_yv12(dst[0], dst[0] + dst_stride[0]*height, dst[0] + dst_stride[0]*height + (dst_stride[0]/2)*height2, dst_stride[0], dst_stride[0]/2, image->y, image->v, image->u, edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP)); return 0; - case XVID_CSP_USER : // u,v swapped + case XVID_CSP_PLANAR: /* YCbCr with arbitrary pointers and different strides for Y and UV */ yv12_to_yv12(dst[0], dst[1], dst[2], - dst_stride[0], dst_stride[1], /* v: dst_stride[2] */ - image->y, image->v, image->u, edged_width, edged_width2, + dst_stride[0], dst_stride[1], /* v: dst_stride[2] not yet supported */ + image->y, image->u, image->v, edged_width, edged_width2, width, height, (csp & XVID_CSP_VFLIP)); return 0; @@ -883,7 +884,7 @@ if (sse==0) return 99.99F; - return 48.131F - 10*(float)log10((float)sse/(float)(pixels)); // log10(255*255)=4.8131 + return 48.131F - 10*(float)log10((float)sse/(float)(pixels)); /* log10(255*255)=4.8131 */ } @@ -907,7 +908,7 @@ return sse; } -/* +#if 0 #include #include @@ -916,7 +917,7 @@ { FILE * f; char hdr[1024]; - + f = fopen(filename, "wb"); if ( f == NULL) { @@ -931,7 +932,7 @@ } -// dump image+edges to yuv pgm files +/* dump image+edges to yuv pgm files */ int image_dump(IMAGE * image, uint32_t edged_width, uint32_t edged_height, char * path, int number) { @@ -954,7 +955,7 @@ return 0; } -*/ +#endif @@ -1094,7 +1095,7 @@ } -/* reduced resolution deblocking filter +/* reduced resolution deblocking filter block = block size (16=rrv, 8=full resolution) flags = XVID_DEC_YDEBLOCK|XVID_DEC_UVDEBLOCK */ @@ -1139,7 +1140,7 @@ for (j = 1; j < mb_height; j++) /* horizontal deblocking */ for (i = 0; i < mb_width; i++) { - if (mbs[(j-1)*mb_stride + i].mode != MODE_NOT_CODED || + if (mbs[(j-1)*mb_stride + i].mode != MODE_NOT_CODED || mbs[(j+0)*mb_stride + i].mode != MODE_NOT_CODED) { hfilter_31(img->u + (j*block - 1)*edged_width2 + i*block, @@ -1149,11 +1150,11 @@ } } - for (j = 0; j < mb_height; j++) /* vertical deblocking */ + for (j = 0; j < mb_height; j++) /* vertical deblocking */ for (i = 1; i < mb_width; i++) { if (mbs[j*mb_stride + i - 1].mode != MODE_NOT_CODED || - mbs[j*mb_stride + i + 0].mode != MODE_NOT_CODED) + mbs[j*mb_stride + i + 0].mode != MODE_NOT_CODED) { vfilter_31(img->u + (j*block)*edged_width2 + i*block - 1, img->u + (j*block)*edged_width2 + i*block + 0,