Camouflage image generation method based on multi-scale generative adversarial network

Abstract

The invention discloses a camouflage image generation method based on a multi-scale generative adversarial network, and the method comprises the following steps: constructing the multi-scale generative adversarial network which comprises a plurality of scales, and each scale comprises a generator, a style conversion network and a discriminator; inputting an initial image of the model, and preprocessing; the generator generates a false image, and inputs the false image and a real image scaled to the same size into the style conversion network and the discriminator for discriminant training; modifying an amplification result of the image generated by the current scale through the image, and then inputting the amplification result into the previous scale; and repeatedly executing the operation of discriminant training, generating an image and inputting the image to the upper-layer scale step until the topmost scale outputs a final camouflage image. According to the method, the single image is trained by constructing the multi-scale adversarial generation network, and the style conversion network is introduced to directionally judge and generate the style of the image, so that the camouflage image is quickly generated by using a small amount of data, and a better camouflage effect is achieved.

Description

technical field [0001] The invention belongs to the field of computer vision, and relates to a camouflage image generation method based on a multi-scale generation confrontation network. Background technique [0002] Image camouflage refers to decorating the foreground with the color and texture of the background, so that the foreground is hidden in the background harmoniously and naturally, while leaving subtle clues different from the background, so that the observer can concentrate on discovering the hidden foreground. Camouflage images have a wide range of applications. In the military field, it can be used for camouflage camouflage of soldiers; in the field of education, there can be images embedded with special content to train children's cognitive ability; in the medical field, pictures of letters confused with the background It can be used to test color blindness; in life, it can be used for artistic creation, entertainment, etc. It is necessary to study the generat...

AI Technical Summary

Problems solved by technology

[0003] At present, the traditional image processing methods are mainly used in image camouflage tasks. Such methods first pass the preprocessing process, artificially design and extract image features, perform image synthesis, and finally pass the post-processing process to obtain a good camouflage effect. The process is complex and computationally complex. large

However, the image camouflage work based on deep learning is still relatively small. The existing work either takes a long time for model training and has no reusability, or the generation speed is fast, but it needs to be driven by large data, and the camouflage image does not have enough data for training.

Images