Abstract:A peak signal-to-noise ratio (PSNR) estimation algorithm that does not use a reference source was developed to address restrictions caused by source signals in calculating the PSNR, which is a widely used index for video quality. The algorithm enables PSNR estimates based on the Laplacian probability density function of the quantized discrete cosine transform (DCT) coefficients extracted from a video bit stream. The algorithm is optimized for the video coding standard H.264 based on its frame structure and its combination of transform and quantization, which accommodates different bit rates and frame types. Tests on various rates and motion videos give average difference of 1dB between the estimated PSNR and the real PSNR for moderate bit rates.

H.264 Coding PSNR Estimation Method Based on Quantized DCT Coefficients