Abstract In this paper, a blind color-image deblurring method based on color image gradients is proposed. The color-image deblurring model is built by introducing color image gradients in the rgb normalized space. The deblurring process is based on alternating iterations of image estimations and blurring kernel estimations. The clear image is obtained by the alternating direction method of multipliers (ADMM), and the blurring kernel is calculated by the half-quadratic penalty method. Experimental results show that the proposed method effectively estimates the blurring kernels. The fidelity of the deblurred images is better, for both motion blurred images and defocused images, than what is achieved with other state-of-the-art blind image deblurring methods. Furthermore, the quality metrics for the proposed method also verify the value of our method compared with other methods.