基于复值小波分解的图象拼合  

Image Mosaics Based on Complex Wavelet Decomposition

作　　者：徐丹[1,2] 鲍歌[1,2] 石教英[1,2] 

机构地区：[1]浙江大学CAD&CG国家重点实验室 [2]香港理工大学电子计算系

出　　处：《软件学报》1998年第9期656-660,共5页Journal of Software

基　　金：国家自然科学基金

摘　　要：图象拼合是一种基于图象的场景编码方法，它被很多基于图象的绘制ＩＢＲ（ｉｍａｇｅｂａｓｅｄｒｅｎｄｅｒｉｎｇ）系统采用，用来建立复杂的虚拟场景表示（例如，３６０°球面和柱面全景图、环境映照及高分辨率图象等）．基于复值小波多分辨率分解（ＣｏｍｐｌｅｘＷａｖｅｌｅｔＭｕｌｔｉｒｅｓｏｌｕｔｉｏｎＤｅｃｏｍｐｏｓｉｔｉｏｎ）提出了一种有效的图象拼合方法，它能同时地、逐步求精地对图象进行匹配和整合．首先，采用复值小波变换不仅可以保证全局优化的结果，还能够满足图象整合的规模不变和平移不变性．其次，基于多分辨率分析可以实现由粗到精的图象匹配和整合，从而使系统与传统的图象整合方法相比具有较高的性能．另外，通过改进图象整合算法中相似距离的测量，降低了计算复杂度．最后，系统可以直接拼合从数字相机捕获的图象，而无需知道相机运动及其他内部参数．Image mosaics is one of the scene encoding approaches and is very popular among many IBR(image based rendering) systems in creating complex virtual environment based on photogeometrics, for example, 360° sphere or cylindrical panoramas, environment maps as well as high resolution images. In this paper, the authors present a robust panoramic image mosaicing scheme which employs complex wavelet pyramid techniques. It addresses the problems of both image matching and registration automatically and simultaneously. Complex wavelet transform guarantees not only a global optimal solution, but also scale and translation invariance for image alignment. The results can be progressively refined on the multiresolution decomposition. This feature guarantees that the scheme has higher performance than the traditional mosaicing techniques. The simplification of similarity measure decreases the complexity of computing. Additionally, the scheme registers images taken directly from digital camera without knowing camera motion and any intrinsic parameters of camera.

关 键 词：图象拼合 复值小波变换 图象匹配 图象整合 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]