摘要
二维小波变换只能很好地分离不连续点,无法最优表示曲线奇异,同时只能获取有限的方向信息,这大大限制了它在图像处理领域的应用。Contourlet变换则结合拉普拉斯金字塔和方向滤波器组,得到多分辨率、局域、多方向的图像表示。由于基于小波变换的多级树集合分裂排序(SPIHT)算法不能有效表达图像的纹理和轮廓信息,因此提出一种基于Contourlet变换和SPIHT算法的彩色图像压缩方法,并应用于医学图像感兴趣区域压缩。首先将彩色图像转换至YIQ彩色空间;然后选取感兴趣区域,对其采用Contourlet变换提取特征信息,并利用SPIHT算法对Contourlet系数优先编码和传输,从而保证感兴趣区域的图像质量和细节信息。对背景区域则采用小波变换,并通过系数截断的方式提高图像压缩比。实验结果表明,所提算法可以较好地保留感兴趣区域的图像特征,大幅度提高背景区域的压缩比,是一种较实用的图像压缩新方法,在医学图像感兴趣区域压缩中效果良好。
Wavelets two-dimension are good at isolating the discontinuities at edge points, but not the smoothness along the contours. In addition, separable wavelets only capture limited directional information, which is restricted in image processing applications. In comparison, contourlet transform combines Laplacian pyramid （LP） with directional filter bank （DFB） to achieve a flexible multi-resolution, local and directional image expansion based on contour segments. A novel image compression method based on contourlet transform and set partitioning in hierarchical trees （SPIHT） algo- rithm was proposed for colorful medical images, because SPIHT algorithm based on wavelet transform can＇t express the texture and contour effectively. Firstly, original RGB image is converted to YIQ color space according to the characteris- tics of human visual system. Secondly, contourlet transform is applied to region of interest （ROI） to capture the main characteristics and then SPIHT algorithm is used to guarantee the compressed image quality and detail. For back ground image,wavelet transform is used to improve compression ratio greatly by wavelet coefficients truncation. Experimental results demonstrate that our algorithm is practical and effective for colorful medical images, which is a good balance for compressed image quality and compression ratio.
出处
《计算机科学》
CSCD
北大核心
2014年第1期303-306,共4页
Computer Science
基金
国家自然科学基金项目(61005054)
南通大学2008年度博士科研启动基金(08B15)资助
