如何在JM8.6程式碼的編碼端提取殘差(為簡便起見,僅以I4x4巨集塊為例)
要在JM8.6程式碼中找出某一資料,不能瞎撞,不能用蠻力,而是要根據H.264的流程來找.比如,要在編碼端找到畫素殘差,就可以按照下面的思路來找. (為簡便起見,僅以I4x4巨集塊為例)
提前說一句:有時候對於同一個量,有好幾個地方多可以提取,所以思路不要太侷限,重要的是知道該量過去從何而來,將來又何去何從. 下面的思路是提取I4x4巨集塊中殘差的思路.(囉嗦一句:如何要提取所有型別巨集塊的殘差,那麼僅在下面這個地方提取則是不夠的)
前面說過,原始的YUV畫素儲存在imgY_org中,而編碼端要對殘差進行變換編碼,而encode_one_macroblock就是編碼一個巨集塊的關鍵函式,所以,如果要在編碼端提取殘差,就可以嘗試在encode_one_macroblock所在的檔案搜尋imgY_org, 一搜索,還真有,但很多,回想一下,殘差是怎麼形成的?很顯然殘差的形成必然涉及到相減,也就是要在程式碼中找“imgY_org[][] - ...”這樣形式的程式碼. 找找,還真有,但是我們需要的“imgY_org[][] - ...”形式的程式碼卻在Mode_Decision_for_4x4IntraBlocks函式中,沒關係的,因為encode_one_macroblock呼叫了Mode_Decision_for_4x4IntraBlocks函式.
提取殘差的程式碼如下:(其中只有部分程式碼用於提取殘差,以I4x4巨集塊為例))
int Mode_Decision_for_4x4IntraBlocks (int b8, int b4, double lambda, int* min_cost) { int ipmode, best_ipmode = 0, i, j, k, x, y, cost, dummy; int c_nz, nonzero = 0, rec4x4[4][4], diff[16]; double rdcost; int block_x = 8*(b8%2)+4*(b4%2); int block_y = 8*(b8/2)+4*(b4/2); int pic_pix_x = img->pix_x+block_x; int pic_pix_y = img->pix_y+block_y; int pic_opix_x = img->opix_x+block_x; int pic_opix_y = img->opix_y+block_y; int pic_block_x = pic_pix_x/4; int pic_block_y = pic_pix_y/4; double min_rdcost = 1e30; int left_available, up_available, all_available; int upMode; int leftMode; int mostProbableMode; static int controlFlag = 0; PixelPos left_block; PixelPos top_block; getLuma4x4Neighbour(img->current_mb_nr, block_x/4, block_y/4, -1, 0, &left_block); getLuma4x4Neighbour(img->current_mb_nr, block_x/4, block_y/4, 0, -1, &top_block); // constrained intra pred if (input->UseConstrainedIntraPred) { left_block.available = left_block.available ? img->intra_block[left_block.mb_addr] : 0; top_block.available = top_block.available ? img->intra_block[top_block.mb_addr] : 0; } upMode = top_block.available ? img->ipredmode[top_block.pos_x ][top_block.pos_y ] : -1; leftMode = left_block.available ? img->ipredmode[left_block.pos_x][left_block.pos_y] : -1; mostProbableMode = (upMode < 0 || leftMode < 0) ? DC_PRED : upMode < leftMode ? upMode : leftMode; *min_cost = (1<<20); //===== INTRA PREDICTION FOR 4x4 BLOCK ===== intrapred_luma (pic_pix_x, pic_pix_y, &left_available, &up_available, &all_available); //===== LOOP OVER ALL 4x4 INTRA PREDICTION MODES ===== for (ipmode=0; ipmode<NO_INTRA_PMODE; ipmode++) { int available_mode = (ipmode==DC_PRED) || ((ipmode==VERT_PRED||ipmode==VERT_LEFT_PRED||ipmode==DIAG_DOWN_LEFT_PRED) && up_available ) || ((ipmode==HOR_PRED||ipmode==HOR_UP_PRED) && left_available ) ||(all_available); if( available_mode) { if (!input->rdopt) { for (k=j=0; j<4; j++) for (i=0; i<4; i++, k++) { diff[k] = imgY_org[pic_opix_y+j][pic_opix_x+i] - img->mprr[ipmode][j][i]; } cost = (ipmode == mostProbableMode) ? 0 : (int)floor(4 * lambda ); cost += SATD (diff, input->hadamard); if (cost < *min_cost) { best_ipmode = ipmode; *min_cost = cost; } } else { // get prediction and prediction error for (j=0; j<4; j++) for (i=0; i<4; i++) { img->mpr[block_x+i][block_y+j] = img->mprr[ipmode][j][i]; img->m7[i][j] = imgY_org[pic_opix_y+j][pic_opix_x+i] - img->mprr[ipmode][j][i]; } //************************************************************** // 下面是提取殘差的程式碼(原始值,預測值和殘差都打印出來) // controlFlag是static int 型別的變數,用於表示巨集塊中的小塊 // 比如controlFlag為2,則表示第一巨集塊的第二個4*4小塊 controlFlag++; if(2 == controlFlag) { printf("****************************\n"); for (j = 0; j < 4; j++) { for (i = 0; i < 4; i++) { printf("%d\t",imgY_org[pic_opix_y+j][pic_opix_x+i]); } printf("\n"); } printf("****************************\n"); for (j = 0; j < 4; j++) { for (i = 0; i < 4; i++) { printf("%d\t",img->mprr[ipmode][j][i]); } printf("\n"); } printf("****************************\n"); for (j=0; j<4; j++) { for (i=0; i<4; i++) { printf("%d\t",img->m7[i][j]); } printf("\n"); } } //************************************************************** //===== store the coding state ===== store_coding_state (cs_cm); // get and check rate-distortion cost if ((rdcost = RDCost_for_4x4IntraBlocks (&c_nz, b8, b4, ipmode, lambda, min_rdcost, mostProbableMode)) < min_rdcost) { //--- set coefficients --- for (j=0; j<2; j++) for (i=0; i<18;i++) cofAC4x4[j][i]=img->cofAC[b8][b4][j][i]; //--- set reconstruction --- for (y=0; y<4; y++) for (x=0; x<4; x++) rec4x4[y][x] = enc_picture->imgY[pic_pix_y+y][pic_pix_x+x]; //--- flag if dct-coefficients must be coded --- nonzero = c_nz; //--- set best mode update minimum cost --- min_rdcost = rdcost; best_ipmode = ipmode; } reset_coding_state (cs_cm); } } } //===== set intra mode prediction ===== img->ipredmode[pic_block_x][pic_block_y] = best_ipmode; img->mb_data[img->current_mb_nr].intra_pred_modes[4*b8+b4] = mostProbableMode == best_ipmode ? -1 : best_ipmode < mostProbableMode ? best_ipmode : best_ipmode-1; if (!input->rdopt) { // get prediction and prediction error for (j=0; j<4; j++) for (i=0; i<4; i++) { img->mpr[block_x+i][block_y+j] = img->mprr[best_ipmode][j][i]; img->m7[i][j] = imgY_org[pic_opix_y+j][pic_opix_x+i] - img->mprr[best_ipmode][j][i]; } nonzero = dct_luma (block_x, block_y, &dummy, 1); } else { //===== restore coefficients ===== for (j=0; j<2; j++) for (i=0; i<18;i++) img->cofAC[b8][b4][j][i]=cofAC4x4[j][i]; //===== restore reconstruction and prediction (needed if single coeffs are removed) ===== for (y=0; y<4; y++) for (x=0; x<4; x++) { enc_picture->imgY[pic_pix_y+y][pic_pix_x+x] = rec4x4[y][x]; img->mpr[block_x+x][block_y+y] = img->mprr[best_ipmode][y][x]; } } return nonzero; }
結果為:
****************************
251 254 254 253
228 205 213 185
197 173 185 136
177 189 198 160
****************************
238 238 238 238
205 205 205 205
167 167 167 167
160 160 160 160
****************************
13 16 16 15
23 0 8 -20
30 6 18 -31
17 29 38 0
三個矩陣分別對應第一幀第一巨集塊的第二個4*4塊的原始值、預測值和殘差,為了驗證上述結果的正確性,可以用H.264visa分析一下,結果如下:
原始值為:(注意: H.264visa也可以讀取原始的YUV資料)
====================== Y Data ======================
+----------------+----------------+----------------+----------------+
| 43,216,254,249,|251,254,254,253,|251,252,254,254,|254,254,254,253,|
| 49,198,193,211,|228,205,213,185,|211,207,186,248,|198,203,208,183,|
| 48,194,177,171,|197,173,185,136,|191,195,138,179,|142,176,177,135,|
| 46,214,225,169,|177,189,198,160,|203,208,177,165,|173,196,191,156,|
+----------------+----------------+----------------+----------------+
| 41,185,208,180,|203,228,226,200,|214,226,225,227,|228,225,224,210,|
| 31,130,173,178,|215,230,221,212,|220,229,227,228,|229,227,226,226,|
| 29,119,194,216,|211,213,219,222,|225,223,220,219,|218,218,218,218,|
| 25,126,219,224,|217,224,227,227,|227,226,225,224,|220,220,221,222,|
+----------------+----------------+----------------+----------------+
| 26,131,215,223,|226,225,225,225,|225,226,223,219,|221,221,219,220,|
| 30,136,216,226,|223,224,225,225,|224,221,217,221,|222,219,220,226,|
| 30,136,216,227,|224,224,225,223,|221,218,221,216,|211,224,224,211,|
| 29,135,217,225,|222,221,222,222,|221,209,181,155,|186,210,186,164,|
+----------------+----------------+----------------+----------------+
| 29,134,216,224,|226,230,230,227,|206,177,146,113,|149,162,147,150,|
| 29,135,219,231,|225,201,190,185,|163,144,153,140,|127,143,165,184,|
| 30,139,210,192,|165,142,134,133,|143,141,129,138,|150,178,201,207,|
| 30,125,166,145,|144,154,132,111,|118,161,175,180,|204,214,213,209,|
+----------------+----------------+----------------+----------------+
預測值為:(注意:H.264visa相當於一個解碼器,解碼端的殘差和編碼端的殘差必定相等)
====================== Y Data ======================
+----------------+----------------+----------------+----------------+
|128,128,128,128,|238,238,238,238,|255,255,255,255,|245,228,210,192,|
|128,128,128,128,|205,205,205,205,|180,180,180,180,|210,192,174,170,|
|128,128,128,128,|167,167,167,167,|137,137,137,137,|174,170,166,166,|
|128,128,128,128,|160,160,160,160,|169,169,169,169,|166,166,166,166,|
+----------------+----------------+----------------+----------------+
|161,161,161,161,|179,189,182,169,|201,201,201,201,|219,219,219,219,|
|161,161,161,161,|184,185,175,169,|214,214,214,214,|227,227,227,227,|
|161,161,161,161,|189,182,169,169,|216,216,216,216,|219,219,219,219,|
|161,161,161,161,|185,175,169,169,|229,229,229,229,|227,227,227,227,|
+----------------+----------------+----------------+----------------+
| 28,127,218,229,|224,224,224,224,|227,227,228,227,|224,224,224,224,|
| 28,127,218,229,|224,224,224,224,|224,225,227,227,|214,214,214,214,|
| 28,127,218,229,|224,224,224,224,|224,224,224,225,|212,212,212,212,|
| 28,127,218,229,|224,224,224,224,|224,224,224,224,|165,165,165,165,|
+----------------+----------------+----------------+----------------+
| 33,130,216,224,|231,223,215,194,|204,182,160,142,|127,133,139,147,|
| 33,130,216,224,|215,194,173,162,|160,142,125,120,|139,147,154,165,|
| 33,130,216,224,|173,162,150,150,|125,120,116,116,|154,165,176,176,|
| 33,130,216,224,|150,150,150,150,|116,116,116,116,|176,176,176,176,|
+----------------+----------------+----------------+----------------+
殘差為:(注意:H.264visa相當於一個解碼器,而解碼端的殘差與編碼端的殘差只是近似,並不相等)
====================== Y Data ======================
+------------------------+------------------------+------------------------+------------------------+
| -78, 88, 132, 110,| 24, 14, 24, 19,| -8, 7, -3, -4,| 16, 31, 48, 51,|
| -81, 63, 67, 77,| 30, 1, 11, -25,| 19, 24, 6, 59,| -12, 20, 32, 10,|
| -83, 62, 48, 39,| 25, 6, 16, -30,| 54, 46, 4, 44,| -30, 7, 5, -32,|
| -80, 87, 93, 32,| 14, 24, 34, 9,| 31, 37, 7, -3,| 6, 28, 21, -9,|
+------------------------+------------------------+------------------------+------------------------+
| -117, 22, 51, 22,| 25, 35, 38, 32,| 18, 18, 18, 18,| 5, 5, 5, 5,|
| -135, -22, 22, 19,| 32, 43, 50, 45,| 13, 13, 13, 13,| 3, 3, 3, 3,|
| -131, -38, 38, 55,| 20, 35, 48, 47,| 3, 3, 3, 3,| -2, -2, -2, -2,|
| -133, -34, 57, 68,| 27, 43, 60, 60,| -2, -2, -2, -2,| -5, -5, -5, -5,|
+------------------------+------------------------+------------------------+------------------------+
| 5, 3, -2, -5,| 0, 0, 0, 0,| -4, -4, -3, -3,| -5, -6, -5, -2,|
| 5, 3, -2, -5,| 0, 0, 0, 0,| -3, -5, -10, -13,| -2, 5, 14, 15,|
| 5, 3, -2, -5,| 0, 0, 0, 0,| -3, -5, -10, -13,| 0, 14, 15, 3,|
| 5, 3, -2, -5,| 0, 0, 0, 0,| 2, -13, -44, -59,| 23, 35, 24, 0,|
+------------------------+------------------------+------------------------+------------------------+
| -4, -4, 0, 4,| 1, 5, 18, 26,| 0, 0, -2, -34,| 25, 20, 8, 3,|
| 2, 7, 11, 10,| 13, 3, 9, 25,| 6, 7, 25, 26,| -9, -3, 9, 15,|
| 0, 6, -8, -28,| -5, -11, -17, -17,| 16, 20, 12, 16,| -5, 4, 22, 31,|
| -6, -5, -39, -74,| -9, 3, -10, -34,| 6, 43, 55, 60,| 33, 33, 35, 35,|
+------------------------+------------------------+------------------------+------------------------+
最後,如果上面程式碼中的if(2 == controlFlag)改為if(1 == controlFlag),那麼結果為:
****************************
43 216 254 249
49 198 193 211
48 194 177 171
46 214 225 169
****************************
128 128 128 128
128 128 128 128
128 128 128 128
128 128 128 128
****************************
-85 88 126 121
-79 70 65 83
-80 66 49 43
-82 86 97 41
與上面的H.264visa的結果進行對比,一切都一目瞭然,便不再贅述了.
P.S. 後來,我發現,上面的過程是有瑕疵的,因為對於一個巨集塊來說,Mode_Decision_for_4x4IntraBlocks函式不止被呼叫16次,所以並不對應16個4*4塊,為什麼會有這種情況呢?跟蹤程式碼後發現,原來對於一個4*4塊而言,要選出最終最合理的幀內預測模式,所要進行的DCT變換次數遠不止16次. 所以如果要真正在編碼端提取畫素殘差,就應該對這麼多次有所選擇, 而JM8.6程式碼必然要作出了這樣的選擇,找找就可以了.