论文部分内容阅读
图像的小波多分辨表征是把图像特征按尺度和方向映射到由小波变换系数构成的金字塔结构的各层中．在此数据结构中，使用不同的基于区域的特征选择方法，实现了对各原始图像的明显特征的选择，包括基于能量判据和基于边缘检测的方法．并且结合这两种方法，综合利用了小波系数的方向信息来进行特征选择．实验结果表明，这些不同的信息融合途径，都能有效地实现基于像素级的多重图像融合，特别是有效地克服由于原始图像的灰度特性和边缘特性不相容对图像融合带来的困难．
Multi-resolution image wavelet representation is the image features by the scale and direction mapping to the wavelet transform coefficients of the pyramid structure of the various layers. In this data structure, different region-based feature selection methods are used to realize the obvious feature selection of each original image, including energy-based criteria and edge-based detection. Combined with these two methods, we make comprehensive use of the direction information of wavelet coefficients to select features. The experimental results show that these different information fusion approaches can all effectively achieve pixel-based multiple image fusion, especially effectively avoiding the difficulties of image fusion due to the incompatibility of the original image’s gray-level features and edge features.