A wavelets based deblocking technique for DCT based compressed materials
Francesco Michielin, G. Calvagno, Piergiorgio Sartor, Oliver Erdler · 2012
In this paper, a new post-processing method operating in the wavelet domain for the suppression of blocking artifacts in DCT compressed images is proposed. The novelties of our method reside in a spatial adaptive equalization in different subbands and in the use of the translation invariant wavelet. The first characteristic is based on the estimation of the local activity in the blocking artifact area, and is suitable for both texture and flat areas. The second feature exploits the local correlation and allows to further improve the quality of the processed images. In particular our method results robust, fast, and suitable for DCT based compressed images at different bit rates. Experimental results show that the proposed method can achieve a significantly better visual quality and also and improvement of the PSNR and SSIM in the output image.