Parent Directory
| 
Revision Log
fixed some bugs (possible overflow, mainly)
/***************************************************************************** * * XVID MPEG-4 VIDEO CODEC * - PSNR-HVS-M plugin: computes the PSNR-HVS-M metric - * * Copyright(C) 2010 Michael Militzer <michael@xvid.org> * * 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. * * 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 * * $Id: plugin_psnrhvsm.c,v 1.2 2010/11/08 20:20:39 Isibaar Exp $ * ****************************************************************************/ /***************************************************************************** * * The PSNR-HVS-M metric is described in the following paper: * * "On between-coefficient contrast masking of DCT basis functions", by * N. Ponomarenko, F. Silvestri, K. Egiazarian, M. Carli, J. Astola, V. Lukin, * in Proceedings of the Third International Workshop on Video Processing and * Quality Metrics for Consumer Electronics VPQM-07, January, 2007, 4 p. * * http://www.ponomarenko.info/psnrhvsm.htm * ****************************************************************************/ #include <stdlib.h> #include <stdio.h> #include <math.h> #include "../portab.h" #include "../xvid.h" #include "../dct/fdct.h" #include "../image/image.h" #include "../utils/mem_transfer.h" #include "../utils/emms.h" typedef struct { uint64_t mse_sum; /* for avrg psnrhvsm */ long frame_cnt; } psnrhvsm_data_t; /* internal plugin data */ static const float CSF_Coeff[64] = { 1.608443f, 2.339554f, 2.573509f, 1.608443f, 1.072295f, 0.643377f, 0.504610f, 0.421887f, 2.144591f, 2.144591f, 1.838221f, 1.354478f, 0.989811f, 0.443708f, 0.428918f, 0.467911f, 1.838221f, 1.979622f, 1.608443f, 1.072295f, 0.643377f, 0.451493f, 0.372972f, 0.459555f, 1.838221f, 1.513829f, 1.169777f, 0.887417f, 0.504610f, 0.295806f, 0.321689f, 0.415082f, 1.429727f, 1.169777f, 0.695543f, 0.459555f, 0.378457f, 0.236102f, 0.249855f, 0.334222f, 1.072295f, 0.735288f, 0.467911f, 0.402111f, 0.317717f, 0.247453f, 0.227744f, 0.279729f, 0.525206f, 0.402111f, 0.329937f, 0.295806f, 0.249855f, 0.212687f, 0.214459f, 0.254803f, 0.357432f, 0.279729f, 0.270896f, 0.262603f, 0.229778f, 0.257351f, 0.249855f, 0.259950f }; static const float Mask_Coeff[64] = { 0.000000f, 0.826446f, 1.000000f, 0.390625f, 0.173611f, 0.062500f, 0.038447f, 0.026874f, 0.694444f, 0.694444f, 0.510204f, 0.277008f, 0.147929f, 0.029727f, 0.027778f, 0.033058f, 0.510204f, 0.591716f, 0.390625f, 0.173611f, 0.062500f, 0.030779f, 0.021004f, 0.031888f, 0.510204f, 0.346021f, 0.206612f, 0.118906f, 0.038447f, 0.013212f, 0.015625f, 0.026015f, 0.308642f, 0.206612f, 0.073046f, 0.031888f, 0.021626f, 0.008417f, 0.009426f, 0.016866f, 0.173611f, 0.081633f, 0.033058f, 0.024414f, 0.015242f, 0.009246f, 0.007831f, 0.011815f, 0.041649f, 0.024414f, 0.016437f, 0.013212f, 0.009426f, 0.006830f, 0.006944f, 0.009803f, 0.019290f, 0.011815f, 0.011080f, 0.010412f, 0.007972f, 0.010000f, 0.009426f, 0.010203f }; #if 0 /* Floating-point implementation */ static uint32_t Calc_MSE_H(int16_t *DCT_A, int16_t *DCT_B, uint8_t *IMG_A, uint8_t *IMG_B, int stride) { int x, y, i, j; uint32_t Global_A, Global_B, Sum_A = 0, Sum_B = 0; uint32_t Local[8] = {0, 0, 0, 0, 0, 0, 0, 0}; uint32_t Local_Square[8] = {0, 0, 0, 0, 0, 0, 0, 0}; float MASK_A = 0.f, MASK_B = 0.f; float Mult1 = 1.f, Mult2 = 1.f; uint32_t MSE_H = 0; /* Step 1: Calculate CSF weighted energy of DCT coefficients */ for (y = 0; y < 8; y++) { for (x = 0; x < 8; x++) { MASK_A += (float)(DCT_A[y*8 + x]*DCT_A[y*8 + x])*Mask_Coeff[y*8 + x]; MASK_B += (float)(DCT_B[y*8 + x]*DCT_B[y*8 + x])*Mask_Coeff[y*8 + x]; } } /* Step 2: Determine local variances compared to entire block variance */ for (y = 0; y < 2; y++) { for (x = 0; x < 2; x++) { for (j = 0; j < 4; j++) { for (i = 0; i < 4; i++) { uint8_t A = IMG_A[(y*4+j)*stride + 4*x + i]; uint8_t B = IMG_B[(y*4+j)*stride + 4*x + i]; Local[y*2 + x] += A; Local[y*2 + x + 4] += B; Local_Square[y*2 + x] += A*A; Local_Square[y*2 + x + 4] += B*B; } } } } Global_A = Local[0] + Local[1] + Local[2] + Local[3]; Global_B = Local[4] + Local[5] + Local[6] + Local[7]; for (i = 0; i < 8; i++) Local[i] = (Local_Square[i]<<4) - (Local[i]*Local[i]); /* 16*Var(Di) */ Local_Square[0] += (Local_Square[1] + Local_Square[2] + Local_Square[3]); Local_Square[4] += (Local_Square[5] + Local_Square[6] + Local_Square[7]); Global_A = (Local_Square[0]<<6) - Global_A*Global_A; /* 64*Var(D) */ Global_B = (Local_Square[4]<<6) - Global_B*Global_B; /* 64*Var(D) */ /* Step 3: Calculate contrast masking threshold */ if (Global_A) Mult1 = (float)(Local[0]+Local[1]+Local[2]+Local[3])/((float)(Global_A)/4.f); if (Global_B) Mult2 = (float)(Local[4]+Local[5]+Local[6]+Local[7])/((float)(Global_B)/4.f); MASK_A = (float)sqrt(MASK_A * Mult1) / 32.f; MASK_B = (float)sqrt(MASK_B * Mult2) / 32.f; if (MASK_B > MASK_A) MASK_A = MASK_B; /* MAX(MASK_A, MASK_B) */ /* Step 4: Calculate MSE of DCT coeffs reduced by masking effect */ for (j = 0; j < 8; j++) { for (i = 0; i < 8; i++) { float u = (float)abs(DCT_A[j*8 + i] - DCT_B[j*8 + i]); if ((i|j)>0) { if (u < (MASK_A / Mask_Coeff[j*8 + i])) u = 0; /* The error is not perceivable */ else u -= (MASK_A / Mask_Coeff[j*8 + i]); } MSE_H += (uint32_t) ((256.f*(u * CSF_Coeff[j*8 + i])*(u * CSF_Coeff[j*8 + i])) + 0.5f); } } return MSE_H; /* Fixed-point value right-shifted by eight */ } #else /* First draft of a fixed-point implementation. Might serve as a template for MMX/SSE code */ static const uint16_t iMask_Coeff[64] = { 0, 59577, 65535, 40959, 27306, 16384, 12850, 10743, 54612, 54612, 46811, 34492, 25206, 11299, 10923, 11915, 46811, 50412, 40959, 27306, 16384, 11497, 9498, 11703, 46811, 38550, 29789, 22598, 12850, 7533, 8192, 10570, 36408, 29789, 17712, 11703, 9637, 6012, 6363, 8511, 27306, 18724, 11915, 10240, 8091, 6302, 5799, 7123, 13374, 10240, 8402, 7533, 6363, 5416, 5461, 6489, 9102, 7123, 6898, 6687, 5851, 6553, 6363, 6620 }; static const uint16_t Inv_iMask_Coeff[64] = { 0, 310, 256, 655, 1475, 4096, 6659, 9526, 369, 369, 502, 924, 1731, 8612, 9216, 7744, 502, 433, 655, 1475, 4096, 8317, 12188, 8028, 502, 740, 1239, 2153, 6659, 19376, 16384, 9840, 829, 1239, 3505, 8028, 11838, 30415, 27159, 15178, 1475, 3136, 7744, 10486, 16796, 27688, 32691, 21667, 6147, 10486, 15575, 19376, 27159, 37482, 36866, 26114, 13271, 21667, 23105, 24587, 32112, 25600, 27159, 25091 }; static const uint16_t iCSF_Coeff[64] = { 1647, 2396, 2635, 1647, 1098, 659, 517, 432, 2196, 2196, 1882, 1387, 1014, 454, 439, 479, 1882, 2027, 1647, 1098, 659, 462, 382, 471, 1882, 1550, 1198, 909, 517, 303, 329, 425, 1464, 1198, 712, 471, 388, 242, 256, 342, 1098, 753, 479, 412, 325, 253, 233, 286, 538, 412, 338, 303, 256, 218, 220, 261, 366, 286, 277, 269, 235, 264, 256, 266 }; static __inline uint32_t isqrt(unsigned long n) { uint32_t c = 0x8000; uint32_t g = 0x8000; for(;;) { if(g*g > n) g ^= c; c >>= 1; if(c == 0) return g; g |= c; } } static uint32_t Calc_MSE_H(int16_t *DCT_A, int16_t *DCT_B, uint8_t *IMG_A, uint8_t *IMG_B, int stride) { int x, y, i, j; uint32_t Global_A, Global_B, Sum_A = 0, Sum_B = 0; uint32_t Local[8] = {0, 0, 0, 0, 0, 0, 0, 0}; uint32_t Local_Square[8] = {0, 0, 0, 0, 0, 0, 0, 0}; uint32_t MASK; uint32_t MSE_H = 0; /* Step 1: Calculate CSF weighted energy of DCT coefficients */ for (y = 0; y < 8; y++) { for (x = 0; x < 8; x++) { uint16_t A = (abs(DCT_A[y*8 + x]) * iMask_Coeff[y*8 + x] + 4096) >> 13; uint16_t B = (abs(DCT_B[y*8 + x]) * iMask_Coeff[y*8 + x] + 4096) >> 13; Sum_A += ((A*A) >> 3); /* PMADDWD */ Sum_B += ((B*B) >> 3); } } /* Step 2: Determine local variances compared to entire block variance */ for (y = 0; y < 2; y++) { for (x = 0; x < 2; x++) { for (j = 0; j < 4; j++) { for (i = 0; i < 4; i++) { uint8_t A = IMG_A[(y*4+j)*stride + 4*x + i]; uint8_t B = IMG_B[(y*4+j)*stride + 4*x + i]; Local[y*2 + x] += A; Local[y*2 + x + 4] += B; Local_Square[y*2 + x] += A*A; Local_Square[y*2 + x + 4] += B*B; } } } } Global_A = Local[0] + Local[1] + Local[2] + Local[3]; Global_B = Local[4] + Local[5] + Local[6] + Local[7]; for (i = 0; i < 8; i++) Local[i] = (Local_Square[i]<<4) - (Local[i]*Local[i]); /* 16*Var(Di) */ Local_Square[0] += (Local_Square[1] + Local_Square[2] + Local_Square[3]); Local_Square[4] += (Local_Square[5] + Local_Square[6] + Local_Square[7]); Global_A = (Local_Square[0]<<6) - Global_A*Global_A; /* 64*Var(D) */ Global_B = (Local_Square[4]<<6) - Global_B*Global_B; /* 64*Var(D) */ /* Step 3: Calculate contrast masking threshold */ { uint32_t MASK_A, MASK_B; uint32_t Mult1 = 64, Mult2 = 64; if (Global_A) Mult1 = ((Local[0]+Local[1]+Local[2]+Local[3])<<8) / Global_A; if (Global_B) Mult2 = ((Local[4]+Local[5]+Local[6]+Local[7])<<8) / Global_B; MASK_A = isqrt(2*Sum_A*Mult1) + 16; MASK_B = isqrt(2*Sum_B*Mult2) + 16; if (MASK_B > MASK_A) /* MAX(MASK_A, MASK_B) */ MASK = (uint32_t) MASK_B; else MASK = (uint32_t) MASK_A; } /* Step 4: Calculate MSE of DCT coeffs reduced by masking effect */ for (j = 0; j < 8; j++) { for (i = 0; i < 8; i++) { uint32_t Thresh = (MASK * Inv_iMask_Coeff[j*8 + i] + 128) >> 8; uint32_t u = abs(DCT_A[j*8 + i] - DCT_B[j*8 + i]) << 10; if ((i|j)>0) { if (u < Thresh) u = 0; /* The error is not perceivable */ else u -= Thresh; } { uint64_t tmp = (u * iCSF_Coeff[j*8 + i] + 512) >> 10; MSE_H += (uint32_t) ((tmp * tmp) >> 12); } } } return MSE_H; } #endif static void psnrhvsm_after(xvid_plg_data_t *data, psnrhvsm_data_t *psnrhvsm) { DECLARE_ALIGNED_MATRIX(DCT, 2, 64, int16_t, CACHE_LINE); uint32_t x, y, stride = data->original.stride[0]; int16_t *DCT_A = &DCT[0], *DCT_B = &DCT[64]; uint8_t *IMG_A = (uint8_t *) data->original.plane[0]; uint8_t *IMG_B = (uint8_t *) data->current.plane[0]; uint64_t MSE_H = 0; for (y = 0; y < data->height; y += 8) { for (x = 0; x < data->width; x += 8) { int offset = y*stride + x; emms(); /* Transfer data */ transfer_8to16copy(DCT_A, IMG_A + offset, stride); transfer_8to16copy(DCT_B, IMG_B + offset, stride); /* Perform DCT */ fdct(DCT_A); fdct(DCT_B); emms(); /* Calculate MSE_H reduced by contrast masking effect */ MSE_H += Calc_MSE_H(DCT_A, DCT_B, IMG_A + offset, IMG_B + offset, stride); } } x = 4*MSE_H / (data->width * data->height); psnrhvsm->mse_sum += x; psnrhvsm->frame_cnt++; printf(" psnrhvsm: %2.2f\n", sse_to_PSNR(x, 1024)); } static int psnrhvsm_create(xvid_plg_create_t *create, void **handle) { psnrhvsm_data_t *psnrhvsm; psnrhvsm = (psnrhvsm_data_t *) malloc(sizeof(psnrhvsm_data_t)); psnrhvsm->mse_sum = 0; psnrhvsm->frame_cnt = 0; *(handle) = (void*) psnrhvsm; return 0; } int xvid_plugin_psnrhvsm(void *handle, int opt, void *param1, void *param2) { switch(opt) { case(XVID_PLG_INFO): ((xvid_plg_info_t *)param1)->flags = XVID_REQORIGINAL; break; case(XVID_PLG_CREATE): psnrhvsm_create((xvid_plg_create_t *)param1,(void **)param2); break; case(XVID_PLG_BEFORE): case(XVID_PLG_FRAME): break; case(XVID_PLG_AFTER): psnrhvsm_after((xvid_plg_data_t *)param1, (psnrhvsm_data_t *)handle); break; case(XVID_PLG_DESTROY): { uint32_t MSE_H; psnrhvsm_data_t *psnrhvsm = (psnrhvsm_data_t *)handle; if (psnrhvsm) { MSE_H = (uint32_t) (psnrhvsm->mse_sum / psnrhvsm->frame_cnt); emms(); printf("Average psnrhvsm: %2.2f\n", sse_to_PSNR(MSE_H, 1024)); free(psnrhvsm); } } break; default: break; } return 0; };
| No admin address has been configured | ViewVC Help |
| Powered by ViewVC 1.0.4 |