Image quality is the basis of image communication and understanding tasks. Due to the blur and noise effects caused by imaging, transmission and other processes, the image quality is degraded. Blind image restoration is widely used to improve image quality, where the main goal is to faithfully estimate the blur kernel and the latent sharp image. In this study, based on experimental observation and research, an adaptively sparse regularized minimization method is originally proposed. The high-order gradients combine with low-order ones to form a hybrid regularization term, and an adaptive operator derived from the image entropy is introduced to maintain a good convergence. Extensive experiments were conducted on different blur kernels and images. Compared with existing state-of-the-art blind deblurring methods, our method demonstrates superiority on the recovery accuracy.