A face reticulate pattern stain removal method based on a multi-task full convolutional neural network

Abstract

The invention requests to protect a face reticulate pattern stain removal method based on a multi-task full convolutional neural network, and the method comprises the following specific steps: 1, taking a clear face image of a public face data set CelebA as a reticulate pattern-free image data set, and making a training set and a verification set; 2, cutting the reticulate pattern image, the realimage and the reticulate pattern binary mask image into image blocks with the size of 64*64; Step 3, obtaining detail layer information of the human face reticulate pattern image as model input by applying guide filtering; 4, designing a full convolutional neural network, and outputting residual error information and binary mask reticulate pattern discrimination information; 5, optimizing the lossof the real image, the reticulate pattern image and the residual error information by using the paired training data, predicting the loss of the reticulate pattern mask and the real reticulate pattern mask label, and training the model; and 6, predicting a test image of a real scene by using the model parameters obtained by training to obtain a predicted decontamination image. According to the invention, a high-quality image is provided for the subsequent operation of the face image.

Description

technical field [0001] The invention belongs to a convolutional neural network and image stain removal technology, in particular to a multi-task end-to-end network stain removal method based on a convolutional neural network. Background technique [0002] Image inpainting is one of the important issues in the field of image algorithms at present, and the effect of image inpainting greatly affects the image recognition results in some scenarios. Image restoration includes image completion, image smudge removal, image super-resolution, etc. For image smudge processing, the existing smudges may affect the semantics represented by the entire image or interfere with the local target of the image. Then, it is necessary to remove the smudges from the image to ensure that the subsequent recognition or detection has a higher accuracy. accuracy. [0003] With the rise of convolutional neural networks in 2012, CNN once again set off a wave of deep learning and artificial intelligence...

AI Technical Summary

Problems solved by technology

The difficulty of stain removal lies in the fuzziness that comes with the removal process and the intractability of strong stains

Images