Abstract: While using compressed sensing theory to process image reconstruction, the thinking based on the block may improve the speed of the reconstruction obviously, but this will bring a strong blocking effects. In order to solve the problem, an image block reconstruction algorithm based on TV norm is proposed, which improves the gradient calculation method based on the entire image, and makes full use of the boundary pixel information that has been reconstructed, thus the image blocking effects are eliminated effectively. Experimental results show that the proposed algorithm can effectively eliminate the image blocking effects, and enhance the reconstructed image’s subjective and objective quality. Compared with TVAL3 algorithm, the PSNR value of reconstructed image can increase 0.84 dB at most, and the reconstructed time can be saved as high as 24.38%. Additionally, the proposed algorithm is especially suitable for the case of low sampling rate.

Key words: compressed sensing, image reconstruction, TV-norm, image blocking

摘要： 利用压缩感知理论进行图像重构时，基于分块思想进行可有效提高重构速度，但同时会带来较强的块效应。为了解决该问题，提出了一种基于TV准则的图像分块重构算法。该算法将基于整幅图像时梯度计算方法进行改进，充分利用已重构块的边界像素信息，从而有效消除了图像的块效应。实验结果表明，提出的算法能够有效消除图像的块效应，提高重构图像的主客观质量，与TVAL3算法相比，重构图像的PSNR值最多提高了0.84 dB，时间最高可节省24.38%，算法尤其适用于低采样率的情况。

关键词: 压缩感知, 图像重构, TV准则, 图像分块