Gain―shape optimized dictionary for matching pursuit video coding

摘要

We show that the vector quantizer (VQ) techniques can be used to optimize a matching pursuit dictionary and improve the PSNR performance at low bitrates. The basis functions are the shape part, and the inner product values are the gain part of the VQ. The performance is evaluated based on a comparison of the PSNR in encoding motion residuals obtained using a matching pursuit coder with a reference dictionary and that obtained from the coder with a gain-shape optimized dictionary.