一种基于皮层变换的图像融合方法

A Cortex Transform Based Image Fusion Method

提出一种基于皮层变换的图像融合方法.该方法模拟生物视觉皮层中简单细胞的响应特性,将单幅图像映射成一系列分辨力及方向各异的子图像,子图像中每个像素代表了视觉皮层中某个神经元的响应.通过简单的展开运算和加法操作就可实现逆变换.仿真实验表明,较之于基于最大绝对系数和最大方向对比度的小波变换融合方法而言,该方法在交叉熵和对比度意义上性能更佳.对于某些源图像而言,其主观效果亦更好.

An image fusion method based on cortex transform is proposed. It simulates the response characteristics of simple cells in biological visual cortex, and decomposes one image into sub-images with different resolutions and orientations in which every pixel represents the response of some neuron in the visual cortex. This method is invertible through simple expansion and additional operations. The results show that when evaluated with the indexes of cross entropy and contrast, it performs better than the wavelet transform based fusion method with the fusion rules of maximum absolute coefficient and maximum directive contl：ast. Furthermore, it is more visually pleasing for some source images.