低比特率Bandelet域图像压缩编码算法研究

Bandelet变换是一种优秀的多尺度几何分析工具,其不仅具有良好的方向性和各向异性,而且能够自适应跟踪图像的几何正则方向.以第二代Bandelet变换为基础,结合Bandelet变换系数分布特性,通过建立最小四叉树、确定最佳量化阈值、构造Bandelet系数自适应扫描方式等措施,提出了一种新的Bandelet变换域图像编码方案.该图像压缩编码方案首先对原始图像进行二维Bandelet变换,并对最低频子带小波系数进行DPCM编码;然后构造最小四叉树并进行四叉树编码,同时进行最佳几何流编码;最后结合最佳量化阈值确定、高频子带自适应扫描等措施,对Bandelet系数进行均匀量化编码.实验结果表明,提出的Bandelet变换域图像编码方法是一种高效的图像压缩算法,不仅其压缩效果明显优于SPIHT、Peyré等图像压缩方案(特别是低比特率下),而且具有比较强的通用性与适应性.

A new image coding algorithm based on second generation Bandelet transform

Finding efficient geometric representation of images is a central issue to improve image com- pression algorithms. The Bandelet transform is an excellent method for representing the geometry of the image, which not only possesses the main features of directionality and anisotropy, but also can take advantage of geometric image regularity by removing the redundancy at different scales and di- rections. In this paper, a new low bit rate image coding algorithm based on the second generation Bandelet transform is proposed. Firstly, the second generation Bandelet transform is performed on o- riginal image, and the lowest frequency wavelet coefficients are encoded by DPCM. Secondly, the smallest quadtrees are constructed and encoded, and the optimal geometric flow is encoded. Finally, the Bandelet coefficients are uniformly quantized by using the optimal quantization threshold and a- daptive scanning order. The experiment results show that the proposed image compression scheme performs better than that of the state-of-art image coders （SPIHT, Peyre＇ method） reported in the literature, especially for low bit-rate and edge/texture image.