Abstract: The medical imaging community has long been interested in brain tumor segmentation, which is an important yet difficult issue. Successful applications of sparse coding and dictionary learning in different vision issues, including picture segmentation, have recently emerged. A superpixel-based framework for automated brain tumor segmentation is presented in this research. It is proposed that the procedures that make up the split and merge technique be reformulated. First, a recursive split is performed; subsequently, after the merge procedure, an image segmentation is acquired. This is possible because the merging is done as a growth process, removing the need for grouping. The usage of a complete quadtree aids in the reformulation process. Because of the differences in nature of the two processes, the region homogeneity in each process is determined using a different predicate. Experiments with a blocks world image and an industrial components image are presented to demonstrate the algorithm's effectiveness.
Keywords: MRI, Brain Tumor, Segmentation, Superpixel, Split and Merge
|
DOI:
10.17148/IJARCCE.2022.11548
