Research On Multi-focus Image Fusion Algorithm

Image Fusion Technology, as an important part of the mufti-sensor information fusion, is a very popular research area in recent years. The actual fusion process can take place at different levels of information representation. A common categorization is to distinguish between pixel, feature and decision level, and the pixel-level image fusion combines multi-source information at the basic level. When screen some object and its aims far different from optical lenses, will suffer from the problem of limited depth of field. In order to solver the problem, can get many images with focus on any aim for some object. These images are called multi-focus image. Multi-focus image fusion technology is propitious to boost the using efficiency of the image information. It can also improve the reliability of the system detecting and recognizing the objects. Because of these merits, it is widely used in these areas such as machine vision, digital camera and object recognition, and so on. This dissertation mainly worked on multi-focus image based on pixel-level fusion.This dissertation laid emphasis on the following points:1. Considering the problems of image fusion performance evaluation, the various methods of objective assessment of image fusion are researched and analyzed systematically. Based on the relationship and individual characteristics of these measures, with the experiment and calculation results, get more appropriate evaluation method for multi-focus image. With the case of standard reference image, root mean square error will be adopted as image fusion objective evaluation method. Without it, mutual information can be adopted as image fusion evaluation method. When compare with the fusion result based on different fusion method, not for the same fusion algorithm, the space frequency or the mean gradient can be adopted as complementary criterion. 2. Through the fundamental research on multi-wavelet, based on good understanding of the fundamental theory and method of image fusion, a new multi-wavelet improved method based on region is presented for combing multi-focus images at pixel-level. This method performs the fusion process finely for both high frequency and low frequency parts of source images. Experimental results show that the proposed method can merge information from source images adequately and improve abilities of information analysis and feature extraction. For further improving the fusion effect, a new improved fusion algorithm is proposed which based on the algorithm that Burt and Kolezynski proposed using the gradient pyramid together with a combination algorithm that is based on an activity and a match menasure and the algorithm that Wilson proposed. Experimental results prove that this method have a better and stabler effect than traditional methods.