Filter evaluation in image deblocking by wavelet thresholding

Abstract

Blockiness occurs when using compression techniques based a block discrete cosine transform (BDCT) at high compression ratios. An approach for reducing the blockiness is based on wavelet thresholding. We evaluate various types of wavelet filters for use in this deblocking technique. The evaluation criteria used include the peak signal-to-noise ratio (PSNR), edge variance (EV) and visual performance. The results show that, among the wavelets studied, the Coiflet-5 filters give best performance for the deblocking purpose.