提出一种用非冗余轮廓波的中低比特率图像质量可伸缩编码算法。该算法采用双正交小波分解和方向滤波器组(DFB)实现图像的非冗余稀疏表示, 不但具有轮廓波对图像中线状奇异性边缘和纹理细节的稀疏表示特点, 而且克服了轮廓波变换系数4/3冗余的缺点。算法中对图像非冗余轮廓波系数各子带系数分布进行统计分析, 通过对变换系数的重新组合, 构造了有利于图像编码的空间方向树结构, 并统计验证了其零树特性, 采用分级树集合分裂和阈值量化达到图像质量可伸缩的嵌入式编码。实验结果表明,其解码算法在中低比特率压缩情况下, 压缩后重构图像的感知质量明显优于小波域SPIHT,JPEG2000编码标准, 峰值信噪比PSNR值与JPEG2000相当, 而图像纹理和边缘细节的视觉效果优于JPEG2000和小波域SPIHT算法。

A novel quality scalable image coding scheme at low-medium based on non-redundancy contourlet transform is proposed. Using the bi-orthogonal wavelet decomposition and directional filter banks, the image non-redundancy sparse representation can be obtained, in which not only it is capable of efficiently characterizing the linear singularity and texture in images but also it can overcome the drawback of 4/3 redundancy in classical contourlet transform. According to the statistical distribution characteristic of coefficients of each subband of non-redundancy contourlet, an elaborated reposition algorithm for the transform coefficients is presented to construct the spatial-orientation tree. Furthermore, the “zero-tree” characteristic of this structure is statistically validated. Then under the principle of set partitioning in hierarchical tree-(SPIHT) algorithm and threshold quantization method, the quality scalable image coding procedure is obtained. Our experimental results demonstrate that at the low to medium rate, the decoding image of the proposed algorithm is competitive to the JPEG2000 in terms of PSNR value, but is visually superior to the wavelet based SPIHT or JPEG2000, especially for those regions with plentiful textures and edge.