深度图像的分块自适应压缩感知

针对基于空域上下采样的深度编码框架中，由边缘信息损失带来的视点绘制质量下降的问题，提出了一种面向视点绘制质量的深度图像分块自适应压缩采样方法。在基于分块压缩感知和光滑Landweber投影重构的BCS_SPL框架下，利用图像块的方差表征其边缘信息，并据此进行自适应采样，以提高深度图像重构和视点合成质量。结果表明，在相同的采样率下，相比上下采样和BCS_SPL方法，本文提出的分块自适应压缩感知方法在绘制视点的PSNR和主观质量上都有提高。

Block-based adaptive compressed sampling of depth image

In the depth coding system based on down/up sampling, the rendered virtual view suffers from image quality loss due to inaccurate edge information of depth images. Based on this fact, this paper proposes a block-based adaptive compressed sampling method of depth image. Based on the BCS_SPL framework combining block-based compressed sampling and smoothed Landweber projection, the proposed method uses variance of blocks to describe edge information for adaptive sampling. The results show that under the same sampling rate, the proposed method achieves improvements both in objective and subjective rendering quality against down/up sampling and BCS_SPL.