Research On Structural Similar Image Fusion Method Based On Image Segmentation

With the development of image processing technology research, image fusion in image processing field has a broad development prospects. Image fusion is to combine the images of the same scene which are attained with the same or different types of imaging sensors into a single one in order to acquire more correct, more accurate, and more complete descriptions and interpretations of the interested objects. With the development of multi-sensor image fusion technology, it has wide applications in the field of medicine diagnosis, military detection, and remote sensing. The fused image makes full use of the input image information to enhance the stability of the image processing system and reduces the redundant information of the input image to enhance the efficiency of image processing system.Fusion method based on image structure similarity in multi-source image fusion can obtain good visual effect, but for the complex image, the fusion effect is not very ideal. This paper introduces the ideas of the image block into image fusion based on structure similarity. After comparing structural similarity in different image block, a set of suitable fusion rules are applied for different image block. The experimental results show that the method of fusion image has better visual effect. Moreover, it provides a feasible fusion schemes for images with complex structures. In this paper, after analyzing the image fusion scheme based on structure similarity and the methods of image block, several image fusion schemes are proposed by combining advantages of two methods. In addition, the objective and subjective effects of fused images are comprehensively evaluated.The main points and innovations in this dissertation are as follows:Firstly, the different regions of the image with abundant contents and complex structures generally have different structures. The structure similarity based on image blocks is cut out for processing the image with complex structures. The processing method can efficiently extract the structure information of the source images, and improve the visual effect of the fused image. on the other hand, Due to the problem that manual threshold of similarity structure arouses overload fusion data and decreases the efficiency of image fusion, a adaptive threshold is attained by computing the standard error of M*N+1 fused image blocks, and the optimal threshold is the threshold corresponding to the maximum standard error. This method can improve the efficiency of image fusion.Secondly, by the idea of block segment of the images and calculation the image clarity of each block, a structure similarity image fusion scheme based image blocks is presented. Firstly, the source images are divided image blocks and the structure similarities of image blocks are computed. Then the image blocks are fused under the threshold of structure similarity attained by the method above mentioned. In addition, how to attain the optimal size of image blocks is also investigated by image experiments. The subjective and objective evaluations of the fused image show that the scheme has the features of simple computation and less time consuming. Therefore, the proposed image scheme improves the efficiency of image fusion.Thirdly, Image fusion method based on wavelet decomposition is better than common method of image fusion, and then an image fusion scheme based on wavelet decomposition and structure similarity with image blocks is proposed. After wavelet decomposition, the high frequency components of the source image contain the detail information, and the low frequency component contains the total structures of the source image. Therefore, the fusion rules is that structure similarity with image blocks is applied to the low frequency component and the selection of wavelet coefficient maximum is applied to the high frequency components. Experimental results show that the scheme can extract more structures from the source images while maintaining the edges of source images. The subjective and objective assessments demonstrate that the fusion method can increase the performance of fused image and attain better visual effects.