Cartoon scene matching system using color and CSIFT feature

Aiming at the cartoon industry which has tremendous applications, a cartoon scene matching system is designed using an improved matching algorithm which fuses color feature and CSIFT (Colored invariant feature transform) feature. The global color statistical feature is obtained by calculating the distribution of three parameters: H, S and I. The CSIFT global feature is calculated based on the color invariance. Two corresponding matching algorithm is brought forward. Finally, the images are matched by fusing color and CSIFT features from two floors: the bottom feature extracting floor and the top feature deciding floor. Experimental results indicate that the two features make up each other, thus it improves matching results of the cartoon scene material greatly. Compared with traditional matching algorithm, the algorithm is robust, and the mean matching precision is up to 80% in the cartoon dataset which has near 500 images.