An Improved Image Retrieval Method Based On Color Spatial Distribution and Shape Curve of Object

Currently, the low-level image features such as color, texture, and shape are widely used for content-based image retrieval. However, it is rather difficult to realize an exact match between two images by using a single feature. To address this problem, a method combining color spatial distribution and shape of the object is presented. In this method, firstly, an improved mean shift algorithm is presented to segment an image into clusters. Secondly, based on these clusters a novel histogram is defined. Thirdly, based on each cluster, the shape curve of an object is extracted. Finally, integrating the information of color and shape of the objects in an image, a novel similarity measure is presented to realize an exact match between two images. Experimental results show that our method can efficiently increase match precision.