Method and system for high-resolution image style transformation with local and global parallel learning

Abstract

The invention discloses a style transformation method based on local and global parallel learning for tens of millions of pixel digital images, which includes the following steps: S1, constructing a stylized model training sample set, including the original image sample set, professional retouchers Manually process the corresponding retouching sample set and the semantic segmentation image sample set corresponding to the original image sample set; S2, compress the original image sample set and the corresponding retouching sample set to obtain a small image training sample set in a small size; S3, Training to obtain a small image stylized model; S4. Based on the training sample set, cut out the original image sample set to obtain corresponding slice pairs, train and record coordinate information, and obtain a slice stylized model; S5. Obtain a fusion model; S6. Joint training steps Three networks in S3‑S5. The invention also discloses a style transformation system based on local and global parallel learning for tens of millions of pixel digital images. The invention has local and global parallel learning, faster processing speed and better effect.

Description

technical field [0001] The invention belongs to the field of image processing, and in particular relates to a style transfer technology for digital single-lens reflex camera imaging. The image with tens of millions of pixels obtained by the digital single-lens reflex camera is passed through a specific sample pair (the original image obtained by the single-lens reflex camera and the corresponding image obtained by the A deep convolutional neural network trained on a stylized image dataset composed of stylized images manually processed by a retoucher to obtain a stylized image, specifically involving a local and global parallel algorithm for digital images with tens of millions of pixels Method and system for learning style transformation. Background technique [0002] The problem to be solved at present is to stylize the photos taken by photographers in some specific layouts or scenes, so as to obtain photos that are more visually aesthetic and stylistic than the original ph...

AI Technical Summary

Problems solved by technology

The problem with this method is that the estimation of its affine bilateral grid is still an approximate estimate. The 16*16*8 affine bilateral grid obtained by the author based on the training picture size of 512 is indeed still accurate when inferring pictures with millions of pixels. Better results can be obtained, but when the size of the picture to be inferred reaches a higher level of tens of millions of pixels, the size of the training picture and the size of the affine bilateral grid need to be increased accordingly, and the training cost will also increase with the In addition, the structure of the model designed by this method is fixed, and the setting of this method forces the model to concentrate most of the parameter fitting pressure on the generation part of the affine bilateral grid. When the effect of the model is not good, it is difficult to carry out two sub-optimization

Images