The Application Research On Medical Image Fusion

Medical image fusion is a subject that includes technology such as signal processing,artificial intelligence technology and so on. This paper is mainly based on the research ofmultimodal medical image fusion. Medical image fusion has lots of methods, and there is noone could be applied to all of the image fusion. Here, according to the research of commonimage fusion methods, we have known that there are some problems of the wavelettransform and traditional PCNN, based on these shortcomings we have done a series ofsimulation and finally raised some new methods as follow to improve the quality of imagefusion:The wavelet transform is a useful tool to process signal time frequency. Due to its maincharacteristic (to highlight the features of some aspects), more and more attention has beenpaid to it in many fields, which has an important application value. In the view of thissituation that in the fusion process of the traditional wavelet, there are problems of lostedges and blurred image texture, so the fusion rule is critical. In the direction of lowfrequency coefficient of wavelet transform, the Tenengrad function take the Sobel to getgradient both for horizontal and vertical, then it is better to keep their original images edgeprofile information. In the direction of the high frequency coefficients of wavelet transform,they are determined by the local variance which is maximal, so that it can not only equalizethe random noise of the image but also avoid too much energy loss of the high frequency. Itwould keep more details information at last. To insure the instability of each image, we haveto check all the coefficients from low frequency and high frequency again. The experimentsresults turned out through this method will have a nice better fusion performance than inother ways.The Pulse Coupled Neural Network (PCNN) has a big difference with the traditionalneural network, it derived from biology. It’s produced according to the synchronous pulsephenomenon of the animals’ cerebral cortex (such as cats and monkey, etc). Due to itsdistinctly superior in some respects, it has been widely used in the areas such as imagesegmentation, image edge detection, image thinning, image recognition, etc. But there arealso problems of large amount of calculation, too many parameters and the user has to set upall of them. So we come up with a solution of these problems as below. Based on theimprovement of two-channel PCNN, a new approach to medical image fusion wasintroduced by using neighborhood spatial frequency inspiring adaptive PCNN, as toimprove the quality of its modal and ability to charge the image of PCNN. The experiments results turned out through this method will have a nice image fusion quality than otherways.