This paper proposes a method of perceptual image hashing to be used in contentbased image retrieval (CBIR). The method is based on an analysis of color and shape/pattern features. The image is first scaled to a standard size, with the pixels classified into 11 categories according to the HSV parameters. From each category, features such as brightness, size and shape are extracted and used to construct the image hash. Since features extracted contain shape and pattern information apart from color, they can effectively represent the image contents. The features are properly weighted to form a hash used in CBIR. Experimental results confirm effectiveness of the proposed method.