基于网格特征跟踪的小波图像压缩方法

WAVELET IMAGE CODING VIA MESH FEATURE TRACKING

用可分离二维小波对图像进行小波变换时，高通滤波器具有水平、竖直和对角三个方向的选择性，因而对其它方向的高频信息（即边缘）的压缩效果不好，在高压缩比的情况下，产生明巴的边缘Ｇｉｂｂｓ现象．本文利用三角形网格对图像进行边缘特征跟踪，有效地去除了这种现象，在至少保持了客观压缩质量（ＰＳＮＲ）的同时，提高了主观压缩质量；并通过确定网格点的特征点匹配领域以及引入弹性网的概念，得到了满足４条原则的网格点坐标的线性代数方程求解算法．

This paper utilizes feature tracking mesh to avoid Gibbs-like phenomena effectively. Both the quality of reconstructed image and the compression ratio are promoted as comparable with the conventional separable wavelet image coding.Meanwhile, by determining the feature neighborhood of mesh points and introducing the concept of spring network, the position of mesh points are obtained by solving a system of linear equations only.