Investigating the PSNR calculation methods for video sequences with source and channel distortions

Measuring video quality is an essential task in video communication, and there exist several methods for its assessment. Among the existing methods, Peak Signal-to-Noise Ratio (PSNR) for its simplicity is the most known and widely used measure. The PSNR of a video sequence can be calculated either as an average of individual frames' PSNR or can be based on the average of individual frames' distortions. In order to demonstrate the accuracy of these two methods, the PSNRs of various video sequences coded at Constant Bit Rate (CBR) and Variable Bit Rate (VBR) have been calculated, and their accuracy is tested against the subjective tests. Experimental results indicate that when there is not much difference between the quality (distortions) of individual frames, like the source coding distortion, the outcome of both methods is similar, but when there is a dominant distortion in some frames, like channel errors, there is a significant difference between the two, and the latter is more correlated with the subjective test results.