A new feature extraction model based on image gradient

A new feature extraction model based on image gradient is proposed. Firstly, image gradient orientation is calculated. Secondly, gradient orientation difference between image pairs is computed at each pixel. Thirdly, the cosine function value of the gradient orientation difference is used as the correlation value. Experimental results show that the correlation calculation model can be used to measure similarity between images effectively. In this way, the proposed method can be used for any existing image recognition method.