提出了一种消除变换编码后解码图象块效应的新方法，它利用邻近象素间存在的相关性，在块间线性预测的基础上定义出对图象相邻子块以及块内的平滑特性产生约束的平滑约束集。在解码端是一个往这些平滑约束集上连续投影的过程，最终得到一幅无论在视觉上还是从信噪比评价的角度来看均具良好效果的解码图象。

In this paper, a new image decoding algorithm to remove the“blocking artifact”appearing with block-transform coding is proposed. In the algorithm, the correlation between adjacent pixels is used to define smooth constraint convex sets. At the decoder, the blocky image is projected onto these convex sets by using the theory of project onto convex sets, and the ultimate result is a smooth image without block artifact. Experiments demonstrate that this new algorithm can not only make the decoded image more smoother but also maintain the details of the image.