The Research Of Multi-focus Image Fusion Based On Fuzzy Sets And Its Fuzzy Measures

Multi-focus image fusion is an effective way to overcome the shortcoming that single image from one sensor can only provide high quality description to a specific area in a give scene.Thus,multi-focus image fusion becomes one of the most important preprocessing technologies in the field of image processing.There are a lot of fuzzy features in the processes of multi-focus imaging because of various factors,such as imaging devices,light conditions,and the inaccuracy and fuzziness of human visual perception.Fuzzy sets theory and its related theories are effective fuzzy information processing tools for dealing with ambiguous and uncertain problems,its natural adaptability and excellent capacity for the fuzzy feature of image show a good prospect in image fusion.But the research of multi-focus image fusion based on fuzzy sets theory and the related theoried are still lack.Fuzzy sets theory and its related theories receive widespread attention of researchers,because of its excellent ability to describe fuzzy information and its foundation position in fuzzy theory.A series of expanded fuzzy sets theories and their information measures have been proposed in succession,but most of the existing measures still have unreasonable decision results when applied in specific issues,such as feature extraction,pattern recognition,and decision making.Therefore more effective measures of fuzzy sets are needed to improve the performance in their applications,which is one of the important research fields for researchers.This thesis focuses on measures of fuzzy sets and multi-focus image fusion,and proposes two new measures with their applications to image fusion.This work can be summarized as the following three aspects:(1)This thesis proposes a transform domain based image fusion method which combines fuzzy sets theory with multi-scale analysis technology to deal with the fuzzy feature in the processes of image fusion.First,SWT is used to decompose source images;second,the membership functions and LSF are utilized to obtain the fuzzy membership matrix of sub-images and their LSF features,successively;and then,the sub-images are fused by fusion rules;finally,the fused image is reconstructed by inverse SWT.Experiments show that this method can obtain satisfactory visual effect.(2)According to the characteristics of multi-focus image,a multi-focus fusion method based on new IFS entropy in NSST domain is presented in this thesis.The IFS entropy is modeled and proposed based on the transformed isosceles triangle fuzzy number of IFS and its complement,and the rationality of the new model is proved by mathematical derivation and experiments.The IFS entropy is used to measure the change degree among pixels in the sub-image decomposed by NSST,and the focusing information of the sub-image is obtained according to the measurement of the IFS entropy.At last,a fusion decision map is obtained to achieve image fusion operation.Experiments not only verify the validity and rationality of the proposed IFS entropy measure,but also show that the proposed multi-focus image fusion method can achieve good fusion effect.(3)Based on the difference of the focues and un-focued ares in multi-focus images,an image focusing detection method based on similarity measure of IFS is proposed for multi-focus images in spatial domain.The similarity measure of IFS is proposed based on the intersections of the transformed symmetrical triangular from IFSs,and its rationality is proved by mathematical derivation and experiments.The new similarity measure is used to get a preliminary fusion decision map by calculating the information of focused area;and then morphological operations are employed to generate the optimized decision map.Besides,an edge reserved method is designed based on Canny edge detection to improve the image fusion performance.The experiments show that the proposed measure is an effective measure which can avoid the problems of most existing measures,and the fusion method is also an effective multi-focus image method,which can achieve better visual effects.The innovation of this thesis lies in proposing the new fuzzy entropy and similarity measure of IFS theory,and applying the proposed theoretical models into multi-focus image fusion.The proposed models can effectively overcome the defects of the existing measures,and provide more high quality fuzzy theory models and ideas for this research field.The corresponding research on image fusion can further expand the application field of fuzzy sets theory,and also provide more technical schemes and theoretical tools for image fusion.