A practical projection-based postprocessing of block-coded images with fast convergence rate

At present, block-transform coding is probably the most popular approach for image compression. However, for the sake of its implementation, an image is partitioned into spatially adjoining blocks which are processed independently without considering inter-block correlation. So, this approach inescapably causes an annoying defect called a blocking artifact. In this letter, in order to reduce a blocking artifact appearing in block-coded images, a new quantization constraint set based on the theory of projection onto convex set (POCS) is proposed. This set can efficiently complement the drawbacks of the projection onto the other constraint sets, particularly the smoothness constraint set. Experimental results, using the proposed quantization constraint set as a substitute for the conventional quantization constraint set, show that the postprocessed images not only converge at a fast rate but also obtain better performance in both objective and subjective quality. Moreover, we know that the postprocessed images maintain the clearness of the decoded image before postprocessing.

目前，块变换编码可能是图像压缩中最常用的方法。然而，为了便于实现，一幅图像被分割成空间上相邻的块，这些块被独立处理，而不考虑块间相关性。因此，这种方法不可避免地会导致一种令人讨厌的缺陷，称为块效应。在本文中，为了减少块编码图像中出现的块效应，基于凸集投影（POCS）理论提出了一种新的量化约束集。该集合可以有效弥补投影到其他约束集（特别是平滑性约束集）的缺陷。实验结果表明，使用所提出的量化约束集替代传统的量化约束集，处理后的图像不仅收敛速度快，而且在客观和主观质量上都能获得更好的性能。此外，我们知道处理后的图像保持了处理前解码图像的清晰度。