Methods and systems for quality controlled encoding

Abstract

This disclosure describes techniques for controlling a perceived quality of multimedia sequences to try to achieve a desired constant perceptual quality regardless of the content of the sequences. In particular, an encoding device may implement quality control techniques to associate a sequence segment with a content “class” based on the content of the segment, determine a perceptual quality metric of the sequence segment, and adjust at least one encoding parameter used to encode the segment is encoded such that for the perceptual quality of the sequence segment converges to the desired quality.

What is claimed is: 
     
         1 . A method for processing multimedia data, the method comprising:
 computing a perceived quality metric for an encoded segment of data associated with digital multimedia data; and   selecting one of a plurality of content classes based on the perceived quality metric and one of at least one encoding parameter used to encode the segment of data and a resultant bitrate of the encoded segment of data.   
     
     
         2 . The method of  claim 1 , further comprising normalizing at least one of the encoding parameters used to encode the segment of data to correspond with encoding parameters used to generate the plurality of content classes. 
     
     
         3 . The method of  claim 2 , wherein the content classes comprise quality-rate curves, the method further comprising:
 computing, for each of the quality-rate curves, a difference between the perceived quality metric and a quality metric on the respective quality-rate curve corresponding to the normalized encoding parameter; and   selecting a certain one of the plurality of quality-rate curves that is associated with a smallest difference of the computed differences.   
     
     
         4 . The method of  claim 1 , wherein the plurality of content classes comprises a plurality of quality-rate curves, and selecting the one of the plurality of content classes includes selecting a certain one of the plurality of quality-rate curves. 
     
     
         5 . The method of  claim 1 , wherein computing the perceived quality metric further includes:
 separating blocks of pixels of frames of data associated with the encoded segment into groups based on at least one difference metric associated with each of the blocks of pixels;   associating quality metric values and weight values with each of the groups of blocks of pixels; and   computing a weighted quality metric for the encoded segment of data based on the quality metric values and weight values associated with of the groups.   
     
     
         6 . The method of  claim 5 , wherein separating the blocks of pixels into groups based on at least one difference metric includes separating the blocks of pixels into groups based on at least one of a sum of absolute difference (SAD), SAD per pixel (SPP), sum of squared differences (SSD), sum of absolute transformed difference (SATD), and sum of squared transformed difference (SSTD). 
     
     
         7 . The method of  claim 5 , wherein separating the blocks of pixels into groups based on at least one difference metric includes:
 separating possible difference metrics into groups, wherein at least a portion of the groups include two or more difference metrics;   pre-computing quality metrics associated with each of the groups, wherein the quality metrics for the groups is equal to an average of quality metrics corresponding to each of the difference metrics associated with the groups; and   pre-computing weights for each of the groups, wherein the weights for each of the groups are computed based on at least a portion of the difference metrics associated with the bins.   
     
     
         8 . The method of  claim 1 , wherein computing the perceived quality metric of the encoded segment of data includes computing an observed peak signal to noise ratio (PSNR) of the encoded segment of data. 
     
     
         9 . An apparatus for processing multimedia data, the apparatus comprising:
 a quality measurement module that computes a perceived quality metric for an encoded segment of data associated with digital multimedia data; and   a class selection module that selects one of a plurality of content classes based on the perceived quality metric and one of at least one encoding parameter used to encode the segment of data and a resultant bitrate of the encoded segment of data.   
     
     
         10 . The apparatus of  claim 9 , further comprising an encoding parameter normalization module that normalizes at least one of the encoding parameters used to encode the segment of data to correspond with encoding parameters used to generate the plurality of content classes. 
     
     
         11 . The apparatus of  claim 10 , wherein the plurality of content classes comprises a plurality of quality-rate curves, and the class selection module also:
 computes, for each of the quality-rate curves, a difference between the perceived quality metric and a quality metric on the respective quality-rate curve corresponding to the normalized encoding parameter; and   selects a certain one of the plurality of quality-rate curves that is associated with a smallest difference of the computed differences.   
     
     
         12 . The apparatus of  claim 9 , wherein the plurality of content classes comprises a plurality of quality-rate curves, and the class selection module selects a certain one of the plurality of quality-rate curves. 
     
     
         13 . The apparatus of  claim 9 , wherein the quality measurement module further:
 separates blocks of pixels of frames of data associated with the segment into groups based on at least one difference metric associated with each of the blocks of pixels;   associates quality metric values and weight values with each of the groups of blocks of pixels; and   computes a weighted quality metric for the segment of data based on the quality metric values and weight values associated with of the groups.   
     
     
         14 . The apparatus of  claim 13 , wherein the quality measurement module also separates the blocks of pixels into groups based on at least one of a sum of absolute difference (SAD), SAD per pixel (SPP), sum of squared differences (SSD), sum of absolute transformed difference (SATD), and sum of squared transformed difference (SSTD). 
     
     
         15 . The apparatus of  claim 13 , wherein the quality measurement module further:
 separates possible difference metrics into groups, wherein at least a portion of the groups include two or more difference metrics;   pre-computes quality metrics associated with each of the groups, wherein the quality metrics for the groups is equal to an average of quality metrics corresponding to each of the difference metrics associated with the groups; and   pre-computes weights for each of the groups, wherein the weights for each of the groups are computed based on at least a portion of the difference metrics associated with the bins.   
     
     
         16 . The apparatus of  claim 9 , wherein the quality measurement module also computes an observed peak signal to noise ratio (PSNR) of the encoded segment of data. 
     
     
         17 . An apparatus for processing multimedia data, the apparatus comprising:
 means for computing a perceived quality metric for an encoded segment of data associated with digital multimedia data; and   means for selecting one of a plurality of content classes based on the perceived quality metric and one of at least one encoding parameter used to encode the segment of data and a resultant bitrate of the encoded segment of data.   
     
     
         18 . The apparatus of  claim 17 , further comprising means for normalizing at least one of the encoding parameters used to encode the segment of data to correspond with encoding parameters used to generate the plurality of content classes. 
     
     
         19 . The apparatus of  claim 18 , wherein the plurality of content classes comprises a plurality of quality-rate curves, and further comprising:
 means for computing, for each of the quality-rate curves, a difference between the perceived quality metric and a quality metric on the respective quality-rate curve corresponding to the normalized encoding parameter; and   means for selecting a certain one of the plurality of quality-rate curves that is associated with a smallest difference of the computed differences.   
     
     
         20 . The apparatus of  claim 17 , wherein the plurality of content classes comprises a plurality of quality-rate curves, and the means for selecting the one of the plurality of content classes includes means for selecting a certain one of the plurality of quality-rate curves. 
     
     
         21 . The apparatus of  claim 17 , wherein the means for computing the perceived quality metric further includes:
 means for separating blocks of pixels of frames of data associated with the segment into groups based on at least one difference metric associated with each of the blocks of pixels;   means for associating quality metric values and weight values with each of the groups of blocks of pixels; and   means for computing a weighted quality metric for the segment of data based on the quality metric values and weight values associated with of the groups.   
     
     
         22 . The apparatus of  claim 21 , wherein the means for separating the blocks of pixels into groups based on at least one difference metric includes means for separating the blocks of pixels into groups based on at least one of a sum of absolute difference (SAD), SAD per pixel (SPP), sum of squared differences (SSD), sum of absolute transformed difference (SATD), and sum of squared transformed difference (SSTD). 
     
     
         23 . The apparatus of  claim 21 , wherein the means for separating the blocks of pixels into groups based on at least one difference metric includes:
 means for separating possible difference metrics into groups, wherein at least a portion of the groups include two or more difference metrics;   means for pre-computes quality metrics associated with each of the groups, wherein the quality metrics for the groups is equal to an average of quality metrics corresponding to each of the difference metrics associated with the groups; and   means for pre-computes weights for each of the groups, wherein the weights for each of the groups are computed based on at least a portion of the difference metrics associated with the bins.   
     
     
         24 . The apparatus of  claim 17 , wherein the means for computing the perceived quality metric for the encoded segment of data includes:
 means for computing an observed peak signal to noise ratio (PSNR) of the encoded segment of data.   
     
     
         25 . A machine readable medium having instructions stored thereon, the stored instructions including one or more portions of code, and being executable on one or more machines, the one or more portions of code comprising:
 code for computing a perceived quality metric for an encoded segment of data associated with digital multimedia data; and   code for selecting one of a plurality of content classes based on the perceived quality metric and one of at least one encoding parameter used to encode the segment of data and a resultant bitrate of the encoded segment of data.   
     
     
         26 . The machine readable medium of  claim 25 , further comprising code for normalizing at least one of the encoding parameters used to encode the segment of data to correspond with encoding parameters used to generate the plurality of content classes. 
     
     
         27 . The machine readable medium of  claim 26 , wherein the plurality of content classes comprises a plurality of quality-rate curves, and further comprising:
 code for computing, for each of the quality-rate curves, a difference between the perceived quality metric and a quality metric on the respective quality-rate curve corresponding to the normalized encoding parameter; and   code for selecting a certain one of the plurality of quality-rate curves that is associated with a smallest difference of the computed differences.   
     
     
         28 . The machine readable medium of  claim 25 , wherein the plurality of content classes comprises a plurality of quality-rate curves, and the code for selecting the one of the plurality of content classes includes code for selecting a certain one of the plurality of quality-rate curves. 
     
     
         29 . The machine readable medium of  claim 25 , wherein the code for computing the perceived quality metric further includes:
 code for separating blocks of pixels of frames of data associated with the segment into groups based on at least one difference metric associated with each of the blocks of pixels;   code for associating quality metric values and weight values with each of the groups of blocks of pixels; and   code for computing a weighted quality metric for the segment of data based on the quality metric values and weight values associated with of the groups.   
     
     
         30 . The machine readable medium of  claim 29 , wherein the code for separating the blocks of pixels into groups based on at least one difference metric includes code for separating the blocks of pixels into groups based on at least one of a sum of absolute difference (SAD), SAD per pixel (SPP), sum of squared differences (SSD), sum of absolute transformed difference (SATD), and sum of squared transformed difference (SSTD). 
     
     
         31 . The machine readable medium of  claim 29 , wherein the code for separating the blocks of pixels into groups based on at least one difference metric includes:
 code for separating possible difference metrics into groups, wherein at least a portion of the groups include two or more difference metrics;   code for pre-computing quality metrics associated with each of the groups, wherein the quality metrics for the groups is equal to an average of quality metrics corresponding to each of the difference metrics associated with the groups; and   code for pre-computing weights for each of the groups, wherein the weights for each of the groups are computed based on at least a portion of the difference metrics associated with the bins.   
     
     
         32 . The machine readable medium of  claim 25 , wherein the code for computing the perceived quality metric for the encoded segment of data includes:
 code for computing an observed peak signal to noise ratio (PSNR) of the encoded segment of data.