Chain Code Approach for Shape based Image Retrieval

Background/Objectives: Shape recognition is the most important issue in image understanding and computer vision. Shape representation is a primary issue in shape recognition. This work aims to image retrieval from image databases based on shape recognition. Methods: The Chain Code Histogram is used to generate a numerical feature vector which is used to describe a given shape; the problems of chain code consist of its sensitivity to rotation, scaling and flipping changes. In this paper, new methods are proposed to solve the problems of image rotation and flipping, by applying the second-moments method to calculating the rotation angle, and analyzing the data of the histogram directions. Findings: The rotation angle of the object and the similarity between the horizontally/ vertically flipping shapes and the original shape. Application: The proposed system works efficiently in recognize both regular and irregular objects for the different categories of the MRI images and results of this research will be useful for the calculate the brain object tilted; research will be extended in near future when we combining this method with color and texture features.