An Efficient Classification of MRI Brain Images and 3D Reconstruction Using Depth Map Estimation

摘要

Digital Image processing is applied under the area of medicine so as to distinguish the ailments in the body of humans. The three dimensional reconstruction (3D) of tumor from the medicinal images is a significant procedure in the area of medicine while it assists the physicians in identification, surgical planning and biological investigation. This article includes two phases namely, i) Classification and ii) 3D reconstruction. Originally the input image is obtained from the MRI database which then undergoes skull stripping is a pre-processing phase for identifying the brain tumor that purges the redundant borough from the image. In the classification phase, the skull Stripped images undergoes segmentation by means of the watershed algorithm so as to identify the segmented tumor. Then from the segmented image the attributes such as shape, intensity and texture are extorted. Subsequently the attributes are lessened via the Principle Component Analysis (PCA). Depending upon the condensed attributes, the probabilistic neural network classifier categorizes the normal and the abnormal (tumor) images. The next phase is the 3D reconstruction phase, we intended the depth assessment for the skull stripped image by means of the guided filter. When the depth is attained, the visual relic of the created left view and right view images yields the ultimate 3D reconstruction outcomes.