The color histogram based image retrieval method is simple and invariant for translation and rotation of the images but losing the spatial information of the color. Recentlymanymethods

such as accumulative histogram

color correlograms

local color histogram

etc

are introduced to improve the color histogrammethod. In this paper

a new content-based color image retrieval method is proposed

in which both the color content and the shape feature of the image have been taken into account. Firstly

based on the special disposal on the HSV color space

an improved accumulative histogram of the hue is calculated as the color feature. To attain the spatial information

H-

S-

and V-component of the image are firstly divided inton×nblocks which are classified into 3 status

flatness

texture and edge status. Then each gray image is translated into a matrix composed of those 3 status values. After that the status matrix is transformed into 1-dimension status sequence

the transition probability matrix of the sequence is calculated as the image’s spatial distribution information. In matching the similarity of the images

the Guassian model is used to normalize the different sub-characters distance. Experiments with different kinds of images indicate that this method is great effective in image’s retrieval.