摘要
图像信号能被分解为低频信号、中频信号和高频信号,基于这一思想将总变差模型滤波器与离散余弦变换、离散条带波变换相结合,提出一种3层的图像压缩算法。首先通过改进的总变差模型滤波器及离散余弦变换得到修正的图像低频信号、中频信号和高频信号。对低频信号即少量的离散余弦系数采用自适应的Huffman编码方法。中频信号含有较丰富的纹理对其进行条带波变换,对其系数进行量化后采用SPIHT算法进行编码。高频信号作为对低频信号的拉普拉斯锐化操作因此只需对滤波器编码即可。Matlab仿真实验验证了方案的可行性,实验中此算法与几种常用的图像变换算法比较,结果表明该算法能达到图像信号更好的稀疏性表示,在保持一定的重构图像质量的前提下可大大提高图像的压缩率。
An efficient three-layered compression algorithm that combines TV model filter and discrete cosine transform （DCT） and discrete Bandelet transform （DBT） based on the idea that an image can be decomposed into the low frequency signal, intermediate signal and high frequency signal was provided. Firstly, the modified image low frequency signal, intermediate signal and high frequency signal were obtained by TVfilter and DCT： The low frequency information that is DCT coefficients is coded by the adaptive Huffman coding. The intermediate signal is discrete Bandelet transformed which includes abundant texture information. The intermediate information that is Bandelet coefficients is quantified and then is coded by SPIHT algorithm. The low frequency signal can be operated by Laplacian sharpening filter and form high frequency signal, therefore it is only need for encoding filter. Matlab simulation validates the feasibility, and experimental results show that the algorithm can obtain a sparse representation of the images signal and can increasing compression ratio ensuring image quality by comparison with some common image transformation methods.
出处
《系统仿真学报》
CAS
CSCD
北大核心
2014年第3期530-536,555,共8页
Journal of System Simulation
基金
国家自然科学基金(60674021)
关键词
总变差滤波器
离散条带波变换
图像压缩
稀疏表示
total variational filter
discrete Bandelet transform
image compression
sparse representation
