Wordart image synthesis system and method based on generative adversarial network

Abstract

The invention requests to protect a wordart image synthesis system and method based on a generative adversarial network. The system comprises an image preprocessing module which is used for preprocessing a wordart image through employing a morphological method, and is used for limiting the font transformation degree during network learning font transformation, astructure transformation module and atexture transformation module. The structure transformation module comprises a generator and a discriminator, the generator is a conversion network and is of a network structure formed by sequentially connecting a plurality of convolution layers, a residual error layer and a deconvolution layer, and the discriminator is provided with two parts, namely a full-connection discriminator and a full-convolution generative adversarial network discriminator; and the texture transformation module is used for adding textures to the image subjected to structure transformation through a cyclic generative adversarial network. Characteristics of the generative adversarial network are utilized to guide generation of wordart pictures with better and better quality. By adopting the system disclosed by the invention, various styles of wordart can be generated, posters of commercial products can be generated through keywords, and the styles and effects of the wordart in the network are extracted.

Description

technical field [0001] The invention belongs to the field of image processing, in particular to image synthesis technology. Background technique [0002] With the development of deep learning in the field of artificial intelligence, the continuous iterative improvement of image style transfer and generative adversarial networks, the research on wordart style transfer has become a research hotspot. The general image style transfer process is as follows: Given a reference image and a target image, the style transfer system can transfer the style of the reference image to the target image to achieve style transfer. Text Effect Style Transfer aims to render text images with style images, resulting in text effects. With the help of generative confrontation networks, it is possible to make analogies to existing complexly designed artistic characters, apply text effects to other characters, and meet different tasks. Posters of commercial products can be generated through the keyw...

AI Technical Summary

Problems solved by technology

The cycle-consistent generative adversarial network used in this technology is very common in the field of style transfer. However, this method is designed for medical images, and there is no structural transformation required for artistic word style image synthesis, because the strokes of the text need to be done by the user. Recognizable, methods using this technology cannot produce clear and recognizable WordArt images

Images