Prashant G. Koparde and Sreelekha Ganapathy
Video compression, Human vision system, Color models, Discrete wavelet transform
This paper presents design and development of a Human Visual System (HVS) based video coder. The major aim of developing such a coder is to highlight the role of perceptual models in improving the reconstructed video quality. The proposed coder incorporates vision models into the coding stream of both the luma and chroma components of the Intra-frames (I-frames) of the input video stream, when represented in the YCbCr space. Unlike the standard video codecs which uses a Discrete Cosine Transform (DCT) based transform coding scheme, a Discrete Wavelet Transform (DWT) based coding is adopted for the compression of I-frames. This not only gives improved compression but also serves as a suitable platform to incorporate the HVS models to improve the quality of the reconstructed video. We call the proposed perceptual video coder a complete video coder since it uses HVS models at all stages of processing. A perceptual metric Structural Similarity Index Measure (SSIM), is used as the cost function for fast motion estimation which improves the speed of processing and prediction. The performance comparison of the proposed and the standard coders are done in terms of latest perceptual measures, such as SSIM and VSNR to highlight the perceptual quality improvements.
Important Links:
Go Back
