Abstract :
From the industrial point of view, image quality is a key-issue. Many post-processing algorithms have been proposed to improve visual quality after the MPEG decoder. Most of them need precise location of the 8×8 grid on which the blocking effect appears. However, in real-life applications, the blocking effect is rarely located on such a basic grid, due to the cascaded bit-rate or format transcoding, rescaling, etc., that occur during acquisition, compression, transmission and display of the video. Consequently, most of these methods see their efficiency largely reduced, or are simply useless. A grid detector is proposed, based on a fine modeling of blocking artifacts in the wavelet domain. It aims at providing essential information to any post-processing algorithm that requires the position of the grid. Several experiments and reliable subjective tests demonstrate the accuracy of the proposed grid detector, and highlight the added value it yields to a post-processing algorithm in terms of visual quality.
Keywords :
data compression; decoding; transform coding; video coding; wavelet transforms; MPEG decoder; arbitrary grid detection; blocking artifacts; cascaded bit-rate; format transcoding; grid detector; image quality; rescaling; wavelet domain; Decoding; Detectors; Displays; Image quality; Robustness; Testing; Transcoding; Transform coding; Video compression; Wavelet domain;