A novel thresholding technique for digital image segmentation

Abstract

Segmentation of region of interest is one among the main image processing step that has been carried on  for any image analysis. A novel thresholding technique for segmenting an input image is  proposed in this paper.  As the first step RGB image is acquired. Taking into account only the green component image,   we convert the image into a logical image and obtain the rank of the resultant image. Using this value as a threshold value we segment the given RGB image. A black and white segmented image is the output image that is generated using this proposed thresholding methodology. $112$ sample images of apples, oranges, bananas and leaves are used for testing the accuracy rate of segmentation. A comparison between the famous Otsu’s thresholding technique and the proposed technique is done and it is found that the accuracy rate of segmentation of the test images using the proposed method is greater than or equal to the output images obtained using Otsu’s histogram thresholding technique.