Efficient shape matching for content-based image retrieval using perceptual grouping

This paper presents an efficient shape matching for content-based image retrieval using perceptual grouping. Humans tend to use high-level concepts in everyday life. The existing computer vision techniques that can automatically extract from an image are mostly low-level features. Object segmentation and recognition is the primary step of computer vision for applying to image retrieval of higher-level image analysis. However, automatic segmentation and recognition of objects via object model is a difficult task without a prior knowledge about the shape of an object. Instead of segmentation and detailed object representation, we proposed an efficient shape matching method which is consisted by two consecutive primitive edge differences (TCPD) into perceptual grouping for image retrieval. Each given image is first segmented into several non-overlapping regions and then scans the different-attributed edge. For speed and efficiency, we extend the TCPD in a query image into several visual-pattern structures that are used to match the images from a large database. Since the basic principle of shape recognition, the proposed method is invariance to translation, rotation, and scale changes. The experimental results demonstrate that the proposed method outperforms the compared method in retrieval accuracy and execution speed.