参数化的Deslauriers-Dubuc插值小波——构造及图像编码性能分析

发布时间：2018-12-12 03:21

【摘要】：研究了采用提升构造具有任意偶数阶消失矩,满足对称性,且仅用一个自由参数表达的Deslauriers-Dubuc(D-D)双正交插值小波.首先,采用多相矩阵理论推导出了此类小波存在的条件;然后,给出了对应小波滤波器和插值小波变换的构造算法.采用算法具体构造了分别具有消失矩对(4,2)、(4,4)、(6,2)以及(6,4)等4类一参数表达的D-D插值小波;最后,以自由参数为自变量,根据编码增益准则,优化设计了4种用于图像编码的插值小波,其滤波器系数全为二进制分数,可实现非乘法运算的离散小波变换(Discrete wavelet transform,DWT).系统分析表明,两种小波的压缩性能超过CDF-9/7小波,对于纹理图像,PSNR增益达到0.44dB,并且计算复杂度可降低17%以上.实验同时表明,新小波的重构图像具有更好的主观可视质量.
[Abstract]:The Deslauriers-Dubuc (D-D) biorthogonal interpolating wavelets with arbitrary even-order vanishing moments, which satisfy symmetry and are expressed only by one free parameter, are studied. Firstly, the existence conditions of this kind of wavelets are derived by using polyphase matrix theory, and then the corresponding wavelet filters and interpolation wavelet transform are constructed. The D-D interpolating wavelets with 4 kinds of parameters, such as (4o 2), (, 4), (, 4), (, 6H 2) and (6H 4), are constructed by using the algorithm. Finally, based on the coding gain criterion, four kinds of interpolated wavelets for image coding are optimally designed with free parameters as independent variables. The filter coefficients are all binary fractions, and the discrete wavelet transform (Discrete wavelet transform,), which can realize non-multiplicative operation, is realized. DWT). The system analysis shows that the compression performance of the two kinds of wavelets is better than that of CDF-9/7 wavelets, and the PSNR gain for texture images is 0.44 dB, and the computational complexity can be reduced by more than 17%. The experiments also show that the reconstructed image of the new wavelet has better subjective visual quality.
【作者单位】：常熟理工学院计算机科学与工程学院;西安交通大学人工智能与机器人研究所;
【基金】：国家重点基础研究发展计划(973计划)(2012CB316400) 国家自然科学基金(61175010;61173130)资助~~
【分类号】：TN919.81

【参考文献】