Multi modal medical image fusion using multi level decomposition with EMGW oriented multi fusion architecture

Abstract

Fusion technology offers more sophisticated details and comprehensive newlineinformation in the field of medical, surveillances, remote sensing and etc. newlineMedical Image Fusion plays a vital role in clinical era which integrates the newlinemulti modal images into a single fused image. These fused images offer more newlineinformation for diagnosing diseases. Decomposition and Fusion are the two newlineimportant functions done in every medical fusion technique. Each newlinedecomposition technique and fusion algorithms are having its own features newlineand drawbacks. In this proposed work, three level of decomposition and three newlinefusion techniques are proposed and obtained target image by combining the newlinefeatures of every decomposing and fusion techniques. The first level newlinedecomposition is done based on the variety of parameters issues by using newlineETV (Enhanced Total Variation) concept. For the small and middle level sub newlinebands, edge details of images are retrieved by using the relationship between newlinethe sub bands and native edge energy is used to form the fusion weight for the newlinecase of large level sub bands. The second level decomposition, Guided-Image newlinefilter is used to decompose the multi modal input images into detailed and newlinesmoothed images with various scales. Next, salient features are retrieved from newlinethose decomposed images using SR (Spectral Residual) and GBVS (Graph newlineBased Visual Saliency). In the third level decomposition, the source images newlineare decomposed into low rank and sparse modules followed by obtaining the newlinecoefficient representations using its corresponding dictionaries. After image newlinepre-processing called image decomposition is done, the multi fusion concept newlineis proposed. For this multi fusion, EMGW (Enhanced Total Variation model, newlineNon-Iterative Multi Exposure and Grey Wolf) optimized fusion architecture is newlineproposed. newline