Intelligent Repair Method of Old Movie Speckle Noise Based on AI Deep Learning

Speckle noise in old movies is caused by erasure or other reasons, which affects the quality of video images. Therefore, an intelligent repair method of speckle noise in old movies based on AI deep learning is proposed. Based on the analysis of the characteristics of speckle noise in old films, the image noise is filtered by bilateral filtering, and the image features are obtained by combining a convolution neural network to obtain the perceptual loss data. On this basis, the jump connection is added to the deep learning network structure of image restoration. Taking the loss function as the training object, the high-quality restoration of speckle noise is realized by optimizing the loss function. The test results show that the design method can ensure that the average PSNR value of the repaired image can reach more than 40 under lower and shorter training iterations, and the effect is obvious.