To solve the problem that image inpainting methods need to provide damaged images and mask images, which limits the use of scenarios, a two-stage image blind inpainting algorithm is proposed based on gated convolutional residual blocks. The algorithm consists of two modules: mask prediction module and image inpainting module. The mask prediction module predicts potential visually inconsistent regions in a given image and outputs a predicted mask; The image inpainting module repairs the damaged area of the image according to the prediction mask and the context of the original image. In order to effectively utilize the contextual information of the image, the algorithm adopts gated convolution in the residual block and introduces a discriminator in the mask prediction module, effectively improving the accuracy of predicting masks. Experimental results on Places2 and CelebA-HQ datasets show that the image repair performance of the proposed algorithm is better than that of the comparison method, and reliable repair images are successfully generated.