基于小波变换的纹理图像融合

Texture image fusion based on wavelet transforms

针对采集到的纹理图像无法呈现纹理物体的整体特征的缺陷,提出了一种新颖的图像融合算法.该融合算法基于纹理图像的大部分纹理信息存在于高频子带中的特点,分别对两幅互补图像进行小波分解,再对低频子带采用平均融合算子处理,然后对高频子带采用高斯-拉普拉斯算子提取局部边缘信息,以作为融合规则,并根据两幅互补图像的相似度对高频子带加以融合.结果表明,该算法通过对多幅互补图像的小波分解图像进行融合,使得融合后的图像内容清晰,纹理信息更加丰富,为后续的缺陷查找步骤提供了准确的依据.

For dealing with the limitation that the acquired image cannot exhibit the whole character ot texture object, a novel image fusing algorithm was proposed. This algorithm was based on the property that a great deal of texture information lying in the high frequency sub-band. Two images with redundancy were decomposed by wavelet transform, and the low frequency sub-band was fused by mean operator. After the local edge information in high frequency sub-band was extracted as fusion rule by Gauss-Laplacian operator, the high frequency sub-band was fused according to the similarity between two images with redundancy. Results show that this fusing algorithm can yield much richer texture information and clear images by fusing many images with redundancy, and provide exact foundation for checking defects.