A method for automatically generating Japanese cartoon portraits

Abstract

The invention discloses a method for automatically generating a portrait of a manga, comprising the steps of: 1. face detection, detection of face feature points, and segmentation of face parts; 2. face recognition in the manga The area is matched with the corresponding face area in the data set. Each face area in the data set has its corresponding manga face area; 3. Generate manga strokes of other parts by using feature points and Japanese manga features; 4. Finally, generate Each part of the face is synthesized into a face caricature according to the geometric relationship of each part of the face. The method of the invention can generate the popular two-dimensional comic image of Riman based on the real human face, which has high entertainment and application value.

Description

technical field [0001] The present invention relates to the technical field of image processing, in particular to a method for automatically generating portraits of manga. Background technique [0002] In recent years, with the digital media and media, Japanese manga and animation have become more and more widely accessible to the Chinese public, and have become more and more popular and sought after by the Chinese public, especially young people. The "two-dimensional" culture has flourished. One of The important reason is that Japanese comics are different from European, American and Chinese comic styles and techniques. However, the drawing of Japanese comics requires a certain amount of drawing talent and painting skills, and it often takes a lot of time for people who have no drawing foundation to draw satisfactory Japanese comics. In addition, in today's mobile terminal photo beautification and various entertainment applications, the cartoonization and sketching of huma...

AI Technical Summary

Problems solved by technology

[0003] In related research work, in the patent CN104077742A, the face photos and sketch samples to be synthesized are divided into non-overlapping pixel blocks, and then the Gabor features of each part of the face are extracted, and the Stein divergence of the pixel blocks is calculated. The method synthesizes each sketch pixel block; in the patent CN104123741A, by calculating the distance transformation value from the non-edge pixel point to the edge pixel point in the binary edge image, the face image in the training library and the same facial organ of the sketch face image to be generated are calculated The distance transformation between them measures the similarity between them, so that the face sketch organ of each part synthesizes the sketch image; but the above methods often retain the local features of the face, but lose the global relationship of the face, and the generated ones are all Sketch effect or cartoon-like face sketch, the style is relatively simple

Images