Research On Multi-focus Image Fusion Based On Block

Multi-focus image fusion based on block mechanism is a special algorithm in the field of multi-focus image fusion. It cuts source images into blocks and compare the resolution of image blocks, then chooses the best image block to generate the fused image. Multi-focus image fusion based on differential evolution treats the image block size as the vector of population and can get the best image block during evolutions. This algorithm is simple and can obtains a good result, but it still has some shortcomings. Firstly, it only generates a offspring population during the evolutions and loses the information of parent population, resulting in a slow convergence speed and a smaller global search scope. Secondly, when the image clarity of the corresponding blocks are the same, it will change the pixels of source images. Thirdly, the parameters are fixed. It will take us more time to adjust parameters in order to obtain a best image block. Lastly, it is easy to fall into local convergence,resulting in a local optimum rather than a global optimum.Considering the shortcomings of the multi-focus image fusion based on differential evolution, several improvements are proposed as follows:Firstly, as an improvement of the multi-focus image fusion based on differential evolution, a new fusion, algorithm is proposed by introducing the twin-generation mechanism and auto-blocking mechanism. The new algorithm generates two offspring populations during the developing process and inherits the major information of parent population. It can expand the global search scope and improve the convergence performance. When the image clary of the corresponding blocks is the same, the new algorithm breaks the image block into smaller blocks and compares the clarity of the smaller blocks,then it can get a better fused image and will not change the pixels of the source images. The experimental results prove that the improved algorithm can get the better fused image than the previous algorithm with a better convergence performance.Secondly, for the problem that the parameters are fixed, another new fusion algorithm based on self-adaptive differential evolution and auto-blocking mechanism is proposed. Based on the improved algorithm, the new algorithm introduces two common parameter adjusting strategy namely liner strategy and random strategy, and for the second one an improvement is designed. The new algorithm can solve the problem that the parameters are fixed and improve the convergence performance by adjusting the parameters automatically. The experimental results prove that the new algorithm has a better performance both on global search scope and convergence performance. It also can get a better fused image. Compared with liner strategy, random strategy is better.