Evaluation of Edge Detection Methods on Different Categories of Images

Object detection in an image is a challenging task. Recent developments in the field of computer vision and machine learning contributes to solving the issue in the field of object detection. Deep learning is one of the recent innovations that selects the feature of an object for evaluation. The shape is the most relevant high-level feature that helps to separate different objects. It can be visualized as a collection of edges and can be defined as a set of contiguous pixel positions where an abrupt change of intensity values occur. Hence, the selection of a better edge detection method for an object category gives higher accuracy in recognition. Our objective in this paper is to compare the various edge detection methods by evaluating the entropy as a measure, to find the best suitable method for each category of objects. Understanding the mechanism behind each of the edge detection algorithms is indispensable to improve the quality of the outcome it produces. Results show the best edge detection for a given category of an image from the Caltech 256 Image Dataset.