变换编码中消除图像“块效应”的优化算法

An Improved Algorithm to Remove the Block Effect in Transform Encoder Images

在基于DCT域内系数量化的压缩算法中,随着码率的降低,量化也越来越粗糙,使重建后的图像产生“块效应”和“振铃现象”等问题。本文提出了一种优化算法来消除这种现象,首先采用块边界滤波削弱相邻块的边界像素突变,然后将图像内的DCT块分为三类,针对不同类型的块所具有的特性采取不同的处理算法,从而在不对图像清晰度产生明显影响的情况下消除“块效应”和“振铃现象”。

In the compression algorithm of coefficient quantization in the DCT domain, as the code rate becomes lower, the quantization becomes more and more rough. Such defects as the block effect and ring effect show up after the image is reconstructed. A mixed algorithm which can weaken these defects is proposed in this paper. First, a method of boundary filtering is used to alleviate the chops between two neighboring DCT blocks, Then, the DCT blocks are classified into three types, and different methods are adopted according to their different characteristics. By this means, these defects are weakened without blurring the image.