Joint Parameter Estimation of Mixed Blur for Image Restoration

The simultaneous existence of motion blur and defocus blur constitutes the mixed blur, which is a great challenge to image deblurring. Therefore, it is necessary to establish the blur model and estimate blur parameters to perform image restoration. In this paper, we propose a joint parameter estimation method for the mix blurred image. For the defocus radius, we derive it based on the correlation between the blurred image and the defocus kernel in spectrum domain. For the motion length, we first roughly estimate it from the spectrum feature. Based on this value, we choose whether to use the correlation in spectrum or make the peak detection in cepstrum to derive the exact motion length. This joint estimation method improves the accuracy when the motion length is small. The experimental results show that the estimation accuracy of motion length and defocus radius is improved, and the estimation range of radius is expanded.