Comparing Subjective Perceived Quality with Objective Video Quality by Content Characteristics and Bit Rates

Different content characteristics and bit rates affecting the perceived quality of video has been verified in this study. We select 93 video sequences encoded/decoded by the Microsoft MPEG-4 software and make use of the principal component analysis (PCA) method to find the principal component for all sequences by motion and texture characteristics. All the sequences are classified into seven groups by the cluster analysis and the discriminant analysis. We examine the different content characteristic sequences with different bit rates by the peak signal noise ratio (PSNR) of the objective video quality evaluation and the double stimulus continuous quality scale (DSCQS) of the subjective quality evaluation. We find when the bit rate for the sequences with the characteristics of higher motion or complex texture needs more than 2,000 kbps, the perceived quality will come to the level of satisfaction. The variance of frame sizes is greater than 2,000 bytes, the bit rate for the sequence with complex texture characteristic needs more than 2,000 kbps. For the types 1, 3, 4 , and 7, when the bit rate is less than 2,000 kbps and the discarded ratio is close to 50% or more than 50%, the subjects still perceive the acceptable satisfaction of video quality. For the types 2 and 5, when the bit rates are greater than 2,000 kbps and the discarded ratio must be controlled below 15%. For type 6, the bit rate needs 2,333 kbps and the discarded ratio must be controlled below 40%.