The Research Of Intracrnialtumor Image Segementation Based On Weighted Aggregation Algorithm

The image segmentation is separated from the image which has a special signifi-cance in the different areas. These regions do not intersect with each other and eachregion should meet the specific area of consistency conditions. Medical image seg-mentation is an integral part of the image segmentation. Intracranial tumor segmenta-tion is an important part of the medical image segmentation, which is very importantto help doctors measure the various parameters of the tumor, quantitative assessmentof brain tumor status and growth process of change, evaluation of treatment effect. Inmedical imaging process, due to the impact of such noise, field shift effect, local bodyeffects etc, brain tumor image acquired usually presents with low contrast，the varia-bility of organizational characteristics，the fuzziness of the boundary between dif-ferent soft tissue or soft tissue lesions，as well as the shape of the structure and dis-tribution of the fine structure (blood vessels, nerves) the complexity of the character-istics which makes it difficult to segment intracranial tumor image. In this paper, allresearches are about the intracranial tumor segmentation based on the weighted ag-gregation algorithm (SWA). Firstly, study of SWA and SWA’s prototype algorithm—Ncut; Then, present the basis of the existing algorithm proposed two improved algo-rithms. This paper’s mainly follows the work and innovation:(1) Verify the importance of Ncut algorithm parameter selection. Understandingdeeply spectral graph theory and normalized cut algorithm based on graph theory andit is important to select parameters K and σ I,σ X to build weight matrix W by ex-perimental verification.(2) SWA intracranial tumors segmentation algorithm based on wavelet transform.This method introduces a wavelet transform, which enhances the high frequency por-tion with the nonlinear high-frequency component image contour compensation andsegments the low frequency component by improved SWA. Firstly, it improves theoriginal seed point selection method In order to enhance the effectiveness of selectseed points. Secondly, it immediately terminate splitting when the best segmentationresults has been extracted(rather than waiting for a pixel image clustering until theend). Finally, the use of seed point region growing algorithm to extract the best seg-mentation results Secondary inverse wavelet transform combine it with enhancedhigh-frequency detail to obtain the final segmentation results. A large number of ex-perimental comparison shows that the accuracy of the improved algorithm is better and the speed of the improved algorithm is more efficient.(3) Intracranial tumors of SWA based on watershed. The most complex of SWAalgorithm is concentrated in the bottom of the pyramid, and the pixel is a one-to-onecorrespondence in the bottom node. In order to improve the efficiency of the SWA isintroduced watershed method instead of the SWA to complete the beginning of thecomposing process. Firstly, watershed algorithm pre-segmentation the image and theimage forms a block, then the SWA method to construct the pyramids rough layer tosegment. For intracranial tumors segmentation, experiments show that the proposedintracranial tumors of SWA based on watershed can effectively improve the process ofcomposition, reduce the running time of the initialization phase, improve the time ef-ficiency of algorithm. But the segmentation effect remains to be further improved.