基于Wedgelet的分层图像压缩

Multi-layered image compression based on Wedgelet

提出了一种分层图像压缩框架:图像=边缘轮廓+纹理。对原图像进行了一种自适应的多尺度Wedgelet分析,抽取并编码了图像的边缘轮廓。基于Wedgelet分析了在残差图像中引入的伪迹所具有的局部振荡特性,采用自适应局部余弦变换分析了以纹理为主要内容的残差图像,在将变换系数重组成与小波系数类似的树形结构后,采用零树编码获得了嵌入式码流。实验结果表明,该算法的重建图像质量优于SPIHT算法,在较好地保留原图像边缘轮廓和有效地减少边缘附近振铃伪迹的同时,较清晰的保留了原图像的纹理特征。

A multi-layered image compression frame was proposed based on modeling images as edge contour ＋ texture. Edge information was represented and coded by a kind of adaptive multiscale Wedgelet analysis. Errors in the Wedgelet approximation and its lossy encoding would create ridge-like artifacts in the residual image. Considering the oscillatory patterns of the artifact ridges, adaptive local cosine transform was adopted to describe the residual texture image. Coefficients were then rearranged to have the wavelet-like zerotree structure, so SPIHT xould be used to generate embedded bitstream. Experimental resuits show that the proposed algorithm outperforms SPIHT visually. In the reconstructed image, the edges of the original image are retained better, and ringing artifacts around the edges are effectively reduced. The textured features are also preserved better.