A Hybrid approach for Image Segmentation

Abstract

This research presents an improved approach for segmentation by using the relationship between pixels which are similar and discontinuous. In similarity relation ,Thresholding segmentation approach is used by finding a good threshold level .Then merging and splitting the pixels of image according to this threshold value. Then Merging the pixels that are grater or equal to the threshold value in gray level and splitting the pixels that are less than threshold value by making the value of these pixels zero so the object will be segmented.In discontinuity relation ,edge deduction is used by splitting the object using sobel operator and merging the lines by using Hough transform. The image in result will be binary image representing the object that was separated from its back ground. Finally the resultant image is an image contains the objects that has the high brightness in similarity thresholding segmentation, but in discontinuity the resulting image is objects contain high frequencies mean that the objects have many edges. Simply it is known that the background has few edges and at many times contains low frequencies so the object here will be segmented by Hough transform segmentation, and also the object that have more brightness will be segmented from the background especially when the background is less brightness.This research is very useful to construct a medicine application applied on MRI ( Magnetic Resonance Imaging ) used to recognize any tumors,….etc.this research is programmed using Visual Basic 6 and shows a good results.