Hair style generation method and device, medium and equipment

Abstract

The invention provides a hair style generation method. The method comprises the following steps: acquiring a to-be-processed input image and a target hair style parameter; rendering a hair style area in the input image by strokes based on the target hair style parameter to obtain a stroke pattern corresponding to the input image; and inputting the stroke pattern into a generative adversarial network model to obtain a target image, the target image being an image after the hairstyle area in the input image is adjusted. According to the invention, stroke rendering is carried out on the input image through the target hair style parameter to generate the stroke pattern, and then model training is carried out on the stroke pattern to generate the high-definition image with a more real hair style. Compared with an existing multi-input-condition hair image generation method, the method has the advantages that the hair structure is more real, and the generated image is higher in definition and more real. In addition, the invention further provides a hair style generation device, a computer readable medium and computer equipment.

Description

technical field [0001] The invention belongs to the field of computer vision, and in particular relates to a method, device, medium and device for generating hairstyles. Background technique [0002] Hair editing is an important topic in the fields of computer vision and computer graphics. Hair is a main feature of an image and is very important for a person's portrait. However, due to the freedom, irregularity and variability of hair, the study of hair is more difficult than that of human face and body. [0003] In recent years, generative adversarial networks have achieved very good results in image generation, and they have also been applied to face generation and portrait attribute control. Face image generation can be controlled through attributes. In portrait generation methods, some studies have involved simple hair attribute control, such as color (black, brown, blonde, etc.) and shape (curly, straight, etc.). However, this is not enough for hair editing. [0004...

AI Technical Summary

Problems solved by technology

Although this method is a breakthrough in the field of hair editing, the images generated by this method are not clear, and the hair structure is often unreasonable

At the same time, due to the use of image restoration technology (inpainting) in this method, the final generated image will have obvious traces of artificial splicing.

Images