基于多方向多尺度变换的图像压缩算法

Image compression algorithms based on multidirectional and multiscale transform

针对小波变换方向选择性差的局限,提出了一种多方向多尺度的的图像变换。圆对称滤波器组首先将图像分解为高频子带和低频子带,然后利用方向滤波器组将高频子带分解为多个方向子带,而对低频子带进行小波变换。多方向多尺度变换能以更稀疏的方式表示图像的边缘和纹理等几何特征,有利于图像压缩。在该变换基础上,结合迭代量化、嵌入式块截断编码(EBCOT)和集合分裂嵌入式块编码(SPECK)构建一种压缩算法。实验结果表明,对于纹理和边缘丰富的图像,压缩算法的性能相对于JPEG2000有明显地提高。

Wavelet transform suffers with the limitation of peor directional selectivity. To overcome this disadvantage, a muhidirectional and multiscale transform was proposde. First the circular symmetric filter bank decomposed the image into high frequency subband and low frequency subband, then the high frequency subband was further decomposed into directional subbands by directional filter bank. The low frequency subband was deoomposed by wavelet transform. The proposed transform could represent the geometrical features such as edges and texture more sparsely, which is of great benefit to image compression. On the base of proposed transform, a compression algorithm combined with iterative quantization, EBOOT and SPECK coder was designed. Experiments show that the proposed algorithms evidendy outperform JPEG2000 for the images with abundant edges and texture.