Method Research On Medical Image Fusion

Medical imaging has become an integral part of modern medicine, but differentmodality medical images have their own advantages and disadvantages. How differentmodality medical image information complementary advantages and comprehensiveexpression for the clinical diagnosis and treatment provide more sufficient, full anatomicaland functional information, it becomes an urgent problem. There is not a universal, accurate,high performance and real-time medical image fusion algorithm. In this paper, the wavelettransform, PCNN and the lifting wavelet transform algorithm in-depth and systematicresearch. Through specific simulation has been a series of valuable conclusions, thecompleted a certain amount of creative work, the specific contents are as follows:In the aspect of different fuzzy MRI medical image fusion, in the view of this situationthat in the fusion process of the traditional wavelet, there are problems of lost edges andblurred image texture, a novel wavelet-based approach for image fusion is presented. Thelow-frequency coefficient of wavelet decomposition based on edge feature, the choicemethod of fused image is better to keep their original images edge profile information. Thehigh-frequency coefficients are selected with local variance which is maximum, so that cancombine with different image characteristics and details effectively; In order to overcomethe presence of noise and control of image instability, all the coefficients are subsequentlyperformed by a window-based consistency verification process. Experimental resultsdemonstrate that the proposed algorithm has better fusion performance than the classicalpyramid and wavelet transform.In the aspect of CT and MRI medical image fusion, the traditional fusion algorithmbased on PCNN, each neuron links with a constant intensity insufficient. A new approach tomedical image fusion was introduced by using neighborhood spatial frequency inspiringadaptive PCNN. First, this algorithm used the Sum-modified-Laplacian of each pixel as thelinking strength of each neuron, so that the linking strength of each pixel can be chosenadaptively; Meanwhile, neighborhood spatial frequency of the pixels were modeled into afeature information to inspire each neuron; Then, the ignition frequency was obtained viathe ignition mapping image generated by the processing of PCNN. The clear objects of eachoriginal image were decided by the compare selection operator and then all of them weremerged into a new clear image. Experimental results demonstrate that the improved PCNNmodel for CT and MRI medical image fusion algorithm is effective. In the aspect of MRA/MRI and MRI-T1/MRI-T2multi-modality medical image fusion,comprehensively utilizing the advantages of Lifting Wavelet Transform (LWT) andimproved Pulse Coupled Neural Network(PCNN), a novel fusion algorithm for medicalimages based on LWT and adaptive PCNN is proposed. Firstly, the source images aredecomposed by LWT, and the sub-band coefficients at different scales are obtained.Secondly, this algorithm uses the local orientation information of wavelet coefficient in eachfrequency domain as the linking strength. After processing PCNN with the adaptive linkingstrength, the ignition frequency is obtained via the ignition mapping image, and then afusion rule in the wavelet domain is given based on the PCNN. Finally, the fusion imagesare reconstructed by wavelet inverse transform. The above method for multi-modal medicalimage fusion has achieved good results, and further proved the feasibility of the method hassome practical value.