<p>Image blind deblurring technology has made significant progress in processing natural images. Saturation is a common phenomenon in blurred images due to low light conditions or long exposure times. Traditional deblurring methods are effective in processing natural images, but they tend to produce ringing effects when restoring saturated images. To address this problem, we propose a robust blind image deblurring algorithm for saturated images. Specifically, we introduce a self-adaptive weight mask by maximum a posteriori that can effectively estimate the saturation characteristics of images. Then we use the alternating direction multiplier method to estimate the fuzzy kernel and the latent image separately. It is worth noting that our algorithm can directly distinguish saturated and non-saturated pixels without complex detection steps. Experiments show that the self-adaptive weight mask effectively reduces the propagation of error in saturated regions, improving the restoration quality of blind deblurring for both synthetic and real-world images.</p>