Image Fusion Algorithms Based On Intersecting Cortical Model

By some fusion algorithm used two different images or multiple images of the same source or on the sensor integrated into a new image is called image fusion. With the development of technology, image fusion technology has qualitative development and improvement, the effect of image fusion has been greatly improved.The main study of this paper is to improve the effect of image fusion, based on the non-subsampled contourlet transform and combined with the third-generation artificial neural network model of intersecting cortical model. The main research works and innovation are as follows:1. The beginning of this article is a brief introduction on the background, significance and the method of image fusion. Then the paper elaborates the course of the wavelet transform image fusion to contourlet transform, then change to the NSCT, but the focuses are the theory of contourlet transformation and NSCT. At last, the article describes in detail to the third generation of pulse neural network model of pulse coupled neural network model and intersecting cortical model.2. The article come up with a further improvement to the traditional intersecting cortical model. The improved intersecting cortical model parameters do not need artificial selection, which can achieve the adaptive selection and meet the real-time image processing. The NSCT transformation has translation invariance properties, therefore, converted parameter selected as the image of multi-scale, multi-directional decomposition tool, and then combined with the improved model for image fusion.3, After NSCT convert of the original image decomposition, the method of the low-frequency portion use the improved sum-modified of Laplace and improved sum-modified energy of Laplace to fusion. The rule of the high frequency parts combine with intersecting cortical model to image fusion.4. Finally, the thesis use three kinds of images which are medical images, infrared and visible images, multi-focus images fusion simulation experiments. Then the thesis though subjective qualitative and other seven objective quantitative indicators to compare with a comprehensive integration of the results of pattern evaluation.Through a large number of simulation experiments show that the algorithm proposed in this paper is much better than others in clarity, the degree of retention of the details of the image, protruding edge information.