Research Of Multi-sensor Decision Level Image Fusion Algorithm Based On D-S Evidence Theory

As the image information acquisition means increasing, the type of image isalso growing. However, there are a lot of redundant information in the acquireddata, in order to get more useful information from the image data, image fusionhas become the focus of many scholars. It is the important embranchments ofdata fusion. In fact, certain algorithms to integrate process two or more images.Its purpose is to synthesis multi-image information in one scene to one imagewhich is more suitable to human observation and computer processing or moreadapt to further image processing, such as target identification. Image fusionreduce uncertainty and redundancy while extracting all the useful informationfrom the source images. Image fusion process is required for differentapplications, medical imaging, remote sensing, machine vision, biometrics andmilitary applications.Since D-S evidence theory allows us to handle uncertain answers from toolsand lack of knowledge about prior probabilities better than the classical Bayesianapproach, In this paper infrared image and visible image are used, based on D-Sevidence theory, to fusion detect targets.This paper analyzes the image features, in the infrared and visible images,the difference between the target and background is gray value. According to thevariety of image features of target and background achieve the classification. Thelocal variance contrast, local variance offset and local entropy three features ofimages as evidence of target detection decision are used. After using the D-Scombination rule to combinate the evidences.Introduce the reason of the evidence conflict of the D-S evidence theory,analysis of existing methods for handing the evidence conflict, based on the basicidea of evidence support degree and evidence distance, introduces the concept ofevidence compatible ratio, presented a new determine method for evidence weight, using D-S evidence theory combination rules for data processing.Compared with other existed methods, this paper considered the conflict andconsistency, calculated the evidence weight objectively.