Evaluation of perceptual visual quality
View Patent ↗A method of evaluating perceptual visual quality (PVQ) of compressed video bit streams in the H.262 or H.264 formats without performing block transforms or prediction decodes I-frames within a group of pictures (GOP). DC and AC coefficients are determined for selected macroblocks within a selected one of the I-frames. Based upon the DC and AC coefficients, the PVQ is calculated.
1. A method of evaluating perceptual video quality (PVQ) for a compressed video bit stream comprising the steps of:
decoding an I-frame within a group of pictures (GOP) within the compressed video bit stream to produce a decoded I-frame;
selecting macroblocks within the decoded I-frame to produce selected macroblocks for analysis, each macroblock having a plurality of blocks;
determining DC and AC coefficients for the selected macroblocks;
calculating the PVQ for the compressed video bit stream as a function of the DC and AC coefficients by:
obtaining a DC difference value between DC coefficients of adjacent blocks in each selected macroblock;
summing the AC coefficients for each block within each selected macroblock to produce AC sum values;
computing a blockiness value for each block as a function of the DC difference values and the AC sum values by multiplying the sum of AC coefficients by qscale and dividing the DC difference value by the sum of AC coefficients times qscale;
computing an average blockiness value per each selected macroblock;
combining the average blockiness values for the selected macroblocks to produce the PVQ indicating a quality of the decoded I-frame based on blockiness values of the selected macroblocks; and
employing the PVQ of the decoded I-frame to predict an overall video quality of the GOP within the video bit stream.
2. The method as recited in claim 1 wherein the selecting step comprises selecting slices within the decoded I-frame as the macroblocks where the compressed video bit stream is an H.262 compressed video bit stream.
3. The method as recited in claim 2 wherein the combining step comprises the steps of:
summing the average blockiness values derived for horizontally adjacent blocks to produce a summed horizontal blockiness value;
summing the average blockiness values derived for vertically adjacent blocks to produce a summed vertical blockiness value;
computing a composite blockiness value as a function of the summed horizontal and vertical blockiness values;
computing an eccentricity value from the summed horizontal and vertical blockiness values;
calculating a composite quality value as a function of the eccentricity value, of the composite blockiness value and of the number of slices within the I-frame being analyzed;
deriving the PVQ from the composite quality value.
4. The method as recited in claim 3 further comprising the step of correcting the summed horizontal and vertical blockiness values as a function of image size prior to the composite blockiness and eccentricity computing steps.
5. The method as recited in claim 2 further comprising the step of gamma correcting the DC difference value prior to input to the blockiness value computing step.
6. The method as recited in claim 1 further comprising the steps of:
repeating the decoding, selecting, determining and calculating steps for a plurality of GOPs to produce a plurality of PVGs; and
averaging the PVGs to determine an overall PVG for the compressed video bit stream.
7. The method of claim 1 , wherein computing the blockiness value further comprises clipping the DC difference and AC sum values to limit particular edges contribution to the PVQ.