基于几何方向的图像压缩算法

摘要

标准二维小波仅沿图像水平和垂直两个方向变换,图像稀疏化表示有待改进完善.文中提出了一种基于方向波变换( Directionlet)的图像压缩算法:通过八叉树分割算法对图像自适应分块；块内最优逼近凡何方向构成Directionlet变换的采样矩阵；块区域进行Directionlet变换,变换系数和方向信息分别编码.沿图像块最优逼近几何方向进行各向异性的Directionlet变换,改善了图像稀疏化效果,提高了编码效率.仿真实验表明,与基于小波的SPIHT和JPEG2000等压缩算法比较,本文方法在数值指标和视觉效果方面均有优势,对几何正则图像,优势更加明显.%The standard wavelet transform fails to provide an efficient representation due to the spatial isotropy of its basis functions built in the horizontal and vertical directions. In this paper, Directionlet based compression scheme was proposed . Using improved octree segmentation to adaptively partition an image into the smaller segments with a dominant direction. Directionlet is applied on a geometrical regular segment, whose optimal geometrical directions constitute the generator matrix of Directionlet transform, coding the coefficients and directional information respectively. Application of Directionlet transform along the optimal geometrical direction per segment efficiently improves sparse representation for image and enhances the coder performance. Experiments demonstrate the superiority of the proposed approach when compared to the wavelet-based SPIHT and JPEG2000 both in terms of objective compression performance measurements as well as visual quality, especially for those images with geometrical regular features.