基于稀疏重建的压缩图像质量提高方法

Compression image quality improvement via sparse reconstruction

为了克服JPEG2000中低码率压缩图像在边缘处表现出的明显振铃效应严重降低了图像的人眼视觉质量的问题,在迭代凸集投影理论基础上,分析了迭代去噪方法恢复图像变换域稀疏数据的机理,提出了JPEG2000压缩图像重建算法.该算法保持压缩图像在小波域中的非零系数不变,通过迭代去噪恢复其他位置的小波系数.结果表明:提出方法有效地增强了JPEG2000压缩图像沿边缘的正则性,提高了图像的主客观质量,图像峰值信噪比平均提高0.52dB.

JPEG2000 compressed image at middle-low bit-rate shows obvious ringing effects,which severely depresses the image visual quality.In order to resolve this problem,on the base of the iterative projection onto convex sets theory,a JPEG2000 image reconstruction method was presented by analysis of the sparse data recovery mechanism in transform-domain using iterated denoising.This algorithm kept the non-zero coefficients in the wavelet domain unchanged and recovered the others coefficients using iterative denoising.Experimental results show that the proposed method is effective to enhance the regularity of JPEG2000 compressed image along the edge and to improve subjective and objective image quality,and the average peak signal to noise ratio is improved by 0.52 dB.