基于显著性信息和方向变换的图像压缩感知

Compressed sensing of images based on saliency information and directional transforms

结合基于图像块的显著性信息随机采样和基于投影Landweber的压缩感知重构算法,提出了一种新的图像压缩感知编码与重构方法。该方法在编码端通过图像显著性信息来分配不同的测量维数以实现测量维数的自适应,在重构端,通过在投影Landweber重构算法中用不同的方向变换来得到重构图像。与同类方法相比,在测量维数相同的前提下,重构图像的峰值信噪比和主观视觉效果都有很大的提高。

A new method of encoding and reconstruction for compressed sensing of image is proposed based on block-based randmn image sampling of saliency information and projected Landweber compressed-sensing reconstruction algorithm. The proposed method achieves the adaptivity of the measurement dimension via saliency information of images in the encoder.In the decoder, the different directional transforms are used in the projected Lane/weber reconstruction algorithm to get the restructured images. Compari- son results with the similar methods show that the proposed method could greatly improve the reconstructed image quality both in PSNR and subjective visual effect on the premise of having the same measuring dimensions.