Abdolah Chalechale - Computer Dept. Faculty of Engineering, Razi University, Iran

This paper presents a novel approach in coin image matching based on middle-level features, instead of low-level features. The approach is developed utilizing edge pixel neighboring histograms for capturing the overall image structure. The closeness of the method to human perceptual judgment is evaluated by a subjective test along with the efficiency in comparison with three other methods known fiom the literature. The query set and the image database used in the tests are scanned, photographed, or collected from the web. Experimental results show a signiJicant supremacy in semantic compatibility over the MPEG-73 edge histogram descriptor (EHD) and the invariant moments method. The proposed method is more eflient than the correlation method which is closer to human perception.

Image analysis, Edge pixels, Coin image recognition, Semantic evaluation