基于小波多尺度分解的肿瘤图像融合

Tumor Image Fusion Based on Wavelet Multiresolution Decomposition

提出了一种基于多尺度小波分解的图像融合方法,该方法利用小波变换对每一图像进行多尺度分解,对各分解层上不同频带的子图像采用不同的融合处理·在高频域内采用模板匹配方法计算出图像区域的统计平均值和方差,从而确定源图像在图像融合中提供信息的比例·在低频域内采用平均算子进行融合,以保留图像的背景信息·最后利用小波逆变换得到融合图像·并用这种方法成功地对肿瘤CT图像进行了融合处理,实验结果表明该融合技术是一种有效的方法,获得的融合图像更适合人们的视觉特性·

A new image fusion technique is put forward using wavelet transform to decompose the multi-resolution signals from each and every image so as to fuse differently the subimages within different bandwidths at every resolution levels. Template matching is used in high-frequency domain to calculate the statistical average and mean square deviation in image display area, thus determining the proportion of information provided by source images in image fusion process. In low-frequency domain the fusion is carried out using averaging operator so as to keep on the background of image. Then, image fusion can be completed by wavelet transform. The technique proposed has successfully been used in CT image fusion for tumor diagnosis, and experimental results revealed that it is an efficient way to provide fused image more adaptable to human vision.