基于灰度纹理信息的图像压缩感知编码与重构

Coding and Reconstruction of Image Compressed Sensing Based on Gray-scale Texture Information

引入了压缩感知(Compressed Sensing,CS)理论,在分析图像DCT系数分布特性的基础上,提出了一种基于灰度纹理信息的压缩采样方法。该方法通过提取图像分块离散余弦变换交流系数的能量,进而对用于对测量过程进行加权修正,充分利用代表图像细节纹理信息的交流分量系数,基于图像轮廓纹理细节信息来分配测量维数,最终实现对不同图像块有区别的压缩采样。比较同类研究结果表明,提出的采样方法在有效减少测量维数或提高重构图像的峰值信噪比和主观视觉效果,以及在降低计算复杂度方面均有更好的表现。

It introduced the CS .（Compressed Sensing ） theory and proposed a novel gray-scale-texture compressive sampling method based on DCT coefficients distribution characteristics for image signals. This method extracts the energy of DCT alternating current coeffi- cient in image block to use for weighted correction in measurement process, makes full use of alternating current comporient coefficient of representing image detail texture information to allocate the measuring dimension based on image contour texture detail information, and ultimately realizes the distinguishing compression sampling for different image blocks. Comparison results with the similar work demon- strate that the proposed compressive sampling method could not only efficiently reduce the computational complexity, but also considera- bly decrease measurement rate and/or enhance the recovery image quality in both PSNR and subjective visual quality.