Medical image segmentation based on improved Ostu algorithm and regional growth algorithm

A new method is proposed for automatic segmentation of edge-blurred medical images, by the which the images are compressed and dissected in multi-resolution through wavelet transform. The images are sharpened by strengthening the extremum of the high-frequents elements at each and every resolution and smoothened by neighboring average and median filtering. Then, the images are segmented automatically using the improved Ostu algorithm(maximization of interclass variance) and the regional growth algorithm through multiseed vote mechanism and, according to the segmenting effect, the threshold is controlled iteratively to evaluate the mean minimum distance with the error found in between the region segmented automatically and that segmented manually. The method proposed has been applied to 30 edge-blurred abdominal MRI images and proves its effectiveness for complex image segmentation.