The Research On Medical Image Fusion Algorithm

With the development of science and technology, medical image fusion technology in the information fusion field has been paid more and more attention to. In order to diagnose, the doctors often allow patients to do all kinds of pathologic examinations, so there are many examination results, It is not correct to determine the patient's condition through only on a single medical image, thus a doctor needs to integrate all the medical images using the naked eyes, obviously it's very difficult to integrate all the medical images by experience, sometimes it's subjective and will result in mistakes, it's not advisable. If all kinds of medical images are effective integrated into a new medical image using some technical methods which reflects the patient's condition, doctors can get more medical information and make an accurate diagnosis or develop the most reasonable treatment program, which has an important effect to medical image fusion technology.The purpose of the medical image fusion technology is to integrate medical images of the same modal or different modes, which can get on with pathological information more accurate, more comprehensive and more reliable information descriptions. Medical image fusion technology involves the artificial intelligence, image processing, pattern recognition and so on. Medical fusion technology is mainly operated in three levels:pixel level fusion methods, feature level fusion methods, symbolic level fusion methods. At present pixel level fusion methods are widely used, which are the foundation of the symbolic level fusion and feature level fusion methods. According to their characteristics, the fusion steps and basic principle, the pixel level fusion methods can be divided into spatial domain fusion methods and transform domain fusion methods. Early, fusion methods are mainly spatial domain fusion methods, which are based on the medical image pixel value. According to some fusion rules, pixel values of different modes of medical images are selected or weighted fused into a new image. With the development of the multi-resolution analysis theory, it provides new opportunities to the image fusion theory, people use multi-resolution analysis theory technology to transform image data from space domain into a transform domain space, fuse them according to transform domain properties of spatial data analysis, finally convert them to their spatial space and obtain the required effects. Along with the information fusion technology has been mature, but medical image fusion technology is just on the start, how to effectively fuse different types of medical images, which is useful.Reading a great number of foreign and domestic best papers, our paper is mainly aimed at medical imaging field of image fusion technology, based on space domain and the transform domain, the pixel level of medical image fusion methods are deep introduced.First, we can study fusion methods based on spatial domain. Fusion methods based on space domain is an important part of the pixel level fusion technology, and also the basis of high level image fusion methods, which are the most widely used fusion methods. This article mainly includes the logic of filtering method, the weighted fusion method, principal component analysis, neural network and so on. Comparing each method, this kind of fusion method is as far as possible to keep the original medical image of raw data, the mathematical theory is also intuitive, algorithm is more simple, but drawbacks of space domain image fusion methods are directly on the image pixels operation, the computation is very large, it's not suitable for real-time processing, the difficulty of this method lies in how to design the most reasonable fusion rules and selecting the fusion coefficients which can reach the best fusion results.Then, we study fusion methods based on transform domain. Image fusion methods based on space domain are direct and simple in mathematics principle, but there are many shortcomings, we need to search more appropriate fusion methods. With the rapid development of the multi-resolution analysis theory and multi-scale analysis theory, the fusion methods based on transform domain gradually replace the fusion methods based on spatial domain. Through image transform techniques, we can transform the images into the transform domain, using the peculiar properties of the images in the transform domain we can fusion the images, using inverse transform method we can get the fused image from transform domain. This article mainly discusses multi-resolution wavelet transform, pyramid transform and multi-scale analysis and so on, comparing the results of two groups of medical image data of each method, fusion methods based on transform domain is better than fusion methods based on space domain. Meanwhile, compared with other fusion methods, the fusion method based on wavelet transform is stable and fusion result is also relatively good, meanwhile multi-scale analysis theory can also obtain good results.At last. the image fusion field of image fusion quality evaluation is an important research direction, image fusion quality assessment has been an image fusion problem, and image fusion technology development also plays a vital role. In the image fusion process, different fusion technology can produce different effects of the fused image, even for the same fusion technology for different kinds of images, fusion results can also be different. Meanwhile, the image fusion technology in different applications, its various parameters of different requirements, which causes the selection of quality evaluation method also will be different, so in order to compare different fusion technology which is more superior, needs to evaluate fused images. This article discusses the information entropy, mutual information, signal-to-noise ratio and average gradient of image quality assessments, these guidelines are all capable of medical image fusion, they evaluate image from each perspective, don't need to rely on human's subjective feelings. These indicators reflect the image fusion method. But in the image fusion performance evaluation process, these indicators have some disadvantages, the single index is only from one prospective to quantify the image. How to seek a complete, objective image fusion performance evaluation system has a long way to go.Along with the science and technology rapid development and computer performance unceasing enhancement, the image fusion technology research has been mature, and will gradually been applied in medical field. Fully use of medical image fusion technology can be integrated all sorts of medical imaging technology advantage, provide abundant information for the disease diagnosis, treatment and prognosis and observing effect which have important significance, has wide application prospects in medical research, will play an important role, this work for subsequent research can make small contributions.