A New Color Pattern Recognition Algorithm Based On Quaternion Correlation

Quaternion correlation, which uses a quaternion to represent the R, G, B value of a pixel in color images, takes advantage of high recognizing precision and strong noise tolerance. However, it can hardly satisfy the speed requirement because of its convolution character. By the Fourier expression of quaternion correlation and the convolution theorem, the color pattern recognition problem is addressed in this paper, to enhance the implementation efficiency. The paper firstly introduces the basic knowledge about quaternion and quaternion correlation. Then, the detailed algorithm is proposed. Finally, we demonstrate the practical implication using example applications for color images, and prove the validity of the algorithm.