Quaternion singular value decomposition approach to color image de-noising

Based on quaternion, an algorithm for color image de-noising has been proposed in this paper. According to the quaternion singular value decomposition theory, in a color image, the singular values on the diagonal matrix, which were obtained through QSVD, represent the color images in different channels. The additive noise of a color image can be eliminated effectively by keeping the proper singular values, that represent normal image signal, and discarding the noise values. Through the color image energy model, we can reconstruct the image singular values and selectively eliminate the singular values which represent the noise. As the result, the proposed algorithm can de-noise color image rapidly, and also it can be implemented easily in practice. The experiment results prove that the algorithm is correct and the energy model is effective.