基于小波变换的图象放大方法再探讨  

Reappraising about Image Magnification Methods Based on Wavelet Transformation

作　　者：刘志刚[1] 刘代志[1] 

机构地区：[1]西安市第二炮兵工程学院602室,西安710025

出　　处：《中国图象图形学报（A辑）》2003年第4期403-408,共6页Journal of Image and Graphics

基　　金：国家自然科学基金 ( 4 99740 2 9) ;国家高等学校骨干教师资助计划资助课题 ( 3 2 5 4)

摘　　要：图象放大 (又称图象变焦、图象重采样 )问题严格地说是一个病态问题 ,根据不同的模型 ,人们先后提出了多种图象放大方法 ,如线性插值、三次样条插值、分形插值以及基于小波变换的方法等 .图象放大问题的焦点是如何在图象放大过程中保持良好的视觉分辨率 ,表现在基于小波变换的图象放大方法中就是如何构造图象高频分量的问题 .针对目前常用的变换域内插方法 ,在一维信号上所作的模拟实验表明 ,该方法并不合理 .对常用的几种图象放大效果评价的方法进行了比较分析 ,结果认为 ,最理想方法还是多人主观评判法 .由于小波高频系数构建问题并未有理想方案 ,因此 ,对小波基函数的选择问题必须作进一步研究 .Strictly speaking, image magnification (image zoom or image resample ) is a morbid problem. Based on various image models, several methods such as bilinear interpolation, bicubic spline interpolation, fractal based interpolation and wavelet based interpolation have been developed. The focus problem of image magnification is how to get good visual resolution. In the methods based on wavelet transformation, it is how to make the detail coefficients of the magnified images. The present method interpolates the coefficients straight in the transformed domain, but the experiments show that the magnified images have bad visual effect. Experiments on one dimension signals in this paper illustrate that method is not reasonable. In addition, two other questions are also discussed: how to estimate the method of image magnification and how to choose the appropriate wavelets in the wavelet based method. Comparing and analyzing show that the best means to estimate the method of image magnification is subjective measure, on the other hand the objective measures mentioned in this paper are merely fit for the traditional interpolation methods. Since there is not a perfect method constructing the detail coefficients of the magnified images at present, how to choose the appropriate wavelets is a new question that should be deeply researched.

关 键 词：小波变换 图象放大方法 视觉分辨率 图象数据处理 变换域 数据内插 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TN911.73[自动化与计算机技术—计算机科学与技术]