Research Of Multi-model Medical Image Fusion Based On Adaptive Cloud Model

With the rapid development of high-dimensional image visualization technology and high-performance computer technology, medical imaging technology has developed to a three-dimensional, dynamic, functional imaging stage. Clinical medicine use several imaging techniques to get different mode of medical imaging, such as magnetic resonance imaging(MRI), computer tomography(CT), positron emission tomography(PET), single photon emission computed tomography(SPECT). Image fusion combains image information which from two or more different modalities, in order to obtain more accurate, comprehensive and reliable image description to the same scene or object. The image fused by different modes can provide richer diagnostic information than single modality image.Currently, there has been many effective fusion methods, each method has its own advantages and disadvantages. Fusion method based on spatial domain deal with gray values directly, its advantage is simple and easy to implement, but the fusion accuracy is often not high. Fusion method based on transform domain transform image on spatial frequency domain firstly, then obtain the fused coefficient according to certain rules, and finally inverse the operation to get the output image. Though fusion method has high fusion accuracy, the fusion process is complex. Cloud model theory is a new reaearch achievement in the field of artificial intelligence, which own both randomness and fuzziness advantages. Integrating cloud model theory into the image fusion achieve the combination of artificial intelligence and image processing effectively, and promote the development of the fusion method based on intelligent domain powerfully.Different from the classical fusion method, the multi-modal medical image fusion method based on adaptive cloud model include two parts: cloud transform of images and the design of cloud inference rules. First, aim at histogram features of the fused image itself, we use high order spline function algorithm to fit the histogram. Then, divide interval according to valley value points of the fitting curve, regard gray values as sample points in each interval, generate three eigenvalues of cloud model through backward cloud algorithm, and obtain cloud model adaptively through positive cloud algorithm according to the above three eigenvalues. Last, design cloud inference rules, as the input gray values for the condition, stimulate X and Y condition cloud generator separately to complete the mapping of gray values on cloud model, output the fused gray value and obtain the fused image finally. Experimental results show that the multi-model medical image fusion method based adaptive cloud model well achieve the fusion process from different modality. The fused image show detail richer and clearer, more obvious characteristics significantly and higher contrast, and get great improvement in not only subjective fusion effect but objective evaluation.