UBC Theses and Dissertations
Blocking artifacts removal of dicrete cosine transform based decompressed images Luo, Ying
Block-based DCT image/video compression methods have been successfully used in image and video compression applications due to their nearly optimal energy compaction property and relative ease of implementation. However, compression distortion becomes significant when these algorithms are used to code images or video under a certain bit rate. The objective of this thesis is to offer post-processing solutions to reduce the block-based DCT image compression artifacts. The emphasis will be placed on the blocking artifacts since the blocking effect is one of the most noticeable degradations of block transform coding. We first analyze the causes and properties of the blocking effect. Then we propose a robust method which detects the locations of the image block boundaries where this effect occurs. The detection results using this method are proved to be accurate at different compression bit rates. Several practical blocking artifacts removal techniques are then reviewed and analyzed. The performance of each technique is studied and evaluated. Finally, a new blocking artifact reduction algorithm which is also effective at very low bit rate is proposed. Our algorithm takes advantage of the fact that the original pixel levels in the same block are of good continuity and we use this property and the correlation between the neighboring blocks to reduce the discontinuity of the pixels across the boundaries where blockiness appear. Our algorithm can highly preserve the high frequency components while smoothing out the boundary discontinuity. Simulation results show that the proposed algorithm significantly reduces the blocking artifacts in both the objective and the subjective measures. It can effectively remove blocking artifacts even at very low bit rates. The amount of computation it requires is also acceptable compared to the iterative blocking artifact removal techniques.
Item Citations and Data