Universal portrait photo cartoon stylizing method

Abstract

The invention discloses a universal portrait photo cartoon stylizing method. The method specifically comprises the following steps: inputting an image S; carrying out cartoon region segmentation on the input image S; dividing the input image S into a portrait feature region A, a portrait non-feature region B and a background region C by using a semantic segmentation algorithm based on deep learning; performing five-sense-organ cartoon special effect algorithm processing on the portrait feature area A to obtain a graph E; carrying out selectable filter algorithm processing on the portrait non-feature area B to obtain a graph F; carrying out selectable filter algorithm processing on the background area C to obtain a graph G; and image fusion: combining the portrait feature area A in the image E, the portrait non-feature area B in the image F and the background area C in the image G to obtain a final effect image D. The beneficial effects of the method are that the method solves a problemthat the output resolution of a pure GAN network is too small; the method is high in universality, and effectively avoids the uncontrollability of the effect and the inconsistency of features.

Description

technical field [0001] The present invention relates to the related technical field of image processing, in particular to a method for stylizing general portrait photos and caricatures. Background technique [0002] At present, various comic-style special effects are often involved in mobile apps or PC image processing software, such as Japanese comics, American comics, Chinese comics, etc. The existing comic processing methods are single, and most of them are special processing for a certain type of comic scenes , does not have universality; at the same time, the caricature effect of the caricature method based on traditional image algorithms (template matching, lineization, etc.) is too poor, and the caricature effect based on the GAN network is too far from the original image of the user (the facial features are not the same), and the output resolution Too small, the result is uncontrollable, and cannot meet the needs of the current market and users. Contents of the inv...

AI Technical Summary

Problems solved by technology

[0002] At present, various comic-style special effects are often involved in mobile apps or PC image processing software, such as Japanese comics, American comics, Chinese comics, etc. The existing comic processing methods are single, and most of them are special processing for a certain type of comic scenes , does not have universality; at the same time, the caricature effect of the caricature method based on traditional image algorithms (template matching, lineization, etc.) is too poor, and the caricature effect based on the GAN network is too far from the original image of the user (the facial features are not the same), and the output resolution Too small, the result is uncontrollable, and cannot meet the needs of the current market and users

Images