Image character editing method of improved FANnet generative network

Abstract

The invention provides an image character editing method of an improved FANnet generative network. The method comprises the following steps: firstly, accurately extracting image characters defined by a user by using an improved adaptive character segmentation model based on an HC saliency detection algorithm; secondly, generating a network according to the FANnet, and generating a target character binary image consistent with the font of the source character; and finally, migrating the color of the source character to the target character through the proposed local color migration model for color complexity discrimination. Therefore, the target editing and modifying character highly consistent with the font structure and color change of the source character is generated, and the purpose of character editing is achieved. Experimental results show that the method is superior to an existing algorithm. For actual scene image characters with relatively complex font structures and color gradient distribution, the invention is also very effective, and has certain theoretical significance and application prospects for image reutilization, automatic error correction of an image character computer, text information restorage and the like.

Description

technical field [0001] The invention belongs to the technical field of image recognition, in particular to an image character editing method for improving the FANnet generation network. Background technique [0002] In today's international society, English characters as an international common language appear in many public places; pinyin characters with the same geometric structure characteristics as English characters are also very important. When these characters appear in the image, especially when the image style is complex, it is difficult to directly edit and modify it. Font style transfer and text generation is an important research field of artificial intelligence. Whether it is a real picture or an electronic rendering, it always contains a lot of text information. These text information can help readers better understand the contextual semantics and scene information in the image. Unlike modifying and editing text in text, when a text in an image is wrong or n...

AI Technical Summary

Problems solved by technology

Its main function is text recognition and extraction. For its editing function, it simply performs image text recognition and erasing and then replaces the modified text, which cannot be guaranteed to be consistent with the original image style.

In 2019, Wu Liang designed three networks: the foreground text migration network, the background erasing network, and the foreground and background fusion network to realize the real-scene image text editing work, but the errors that occur in each network during the editing process will be in the next network Accumulate and expend costs

In 2020, Prasun Roy and others proposed the CNN-based text generation network (FANnet) for the first time, realizing the STEFFAN model for text editing in real-scene images, but because it cannot recognize real scenes with complex font structures or color features well in the source text extraction stage Image characters, so the accuracy of FANnet is not high, and the editing of image characters cannot be well realized

Images