Unlock instant, AI-driven research and patent intelligence for your innovation.

A style transfer method for high-resolution images based on random rearrangement of image blocks

A high-resolution image and image block technology, applied in the field of image processing, can solve the problems of unsatisfactory style transfer at the border splicing, unsatisfactory overall effect of style transfer, unnatural splicing transition, etc., to eliminate splicing traces and noise Color blocks, enhance the overall effect, splicing transition natural and beautiful effect

Active Publication Date: 2020-10-16
ZHEJIANG UNIVERSITY OF SCIENCE AND TECHNOLOGY
View PDF5 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

This method can support high-resolution image style migration without changing the hardware configuration, and is universal. The division of image data blocks in this method is the key. Migration will lead to unsatisfactory overall effect of style migration, and there are problems of unsatisfactory style migration and unnatural splicing transition at the border splicing

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • A style transfer method for high-resolution images based on random rearrangement of image blocks
  • A style transfer method for high-resolution images based on random rearrangement of image blocks
  • A style transfer method for high-resolution images based on random rearrangement of image blocks

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0028] Embodiment 1: A high-resolution image style migration method based on random rearrangement of image blocks. The hardware in this embodiment adopts Intel Xeon E5-2620 CPU, equipped with a NVIDIA Tesla K80; the operating system uses CentOs7.4.1708 operating system, Based on the open source deep learning framework TensorFlow, the version is 1.8.0, and the CUDA version is 9.0.176, including an image x with a length and width of W×H. The style picture, the pixel size is not limited; the steps are as follows figure 1 Proceed as shown:

[0029] a. If figure 1 Steps 1-2 in and figure 2 As shown, the image x is all divided into side length w basic The basic square of the image x, the edge part that cannot be completely separated in the image x expands outwards to make up the width of the basic square, and the length of the new image x expands to The width of the new image x is expanded to Such as figure 1 Step 3 in and image 3 As shown, fill the new image x evenly aro...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention discloses a high-resolution image style transfer method based on random rearrangement of image blocks. Firstly, the image is enlarged, and then a plurality of image squares are generated in order by using the enlarged image, and numbered in order, and then random functions are used to The image squares are arranged, and then multiple image squares are spliced ​​into sub-images that can be processed by the computer according to the upper limit resolution of the computer, and then the individual sub-images are sequentially processed by image grid migration, and the stylized sub-images are cut. The image squares are then arranged and spliced ​​according to the sequence numbers, and the overlapping areas between adjacent image squares are merged. Finally, the expanded width and filled width of the edge of the original image are removed to restore it to the original size, and finally the finished product is obtained. The invention can make the splicing and transition of the boundary after image style transfer natural and beautiful, and can effectively improve the overall effect of style transfer, and has the characteristics of strong applicability.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a high-resolution image style transfer method based on random rearrangement of image blocks. Background technique [0002] Image style transfer is a long-term research topic. Before the emergence of neural network style transfer (NST), related research in computer graphics has been extended to the field of non-photorealistic rendering (NPR). In the field of computer vision, style transfer is usually considered as a problem of texture synthesis, and it is widely used in social networking, user-assisted creation, entertainment application creation and other fields. In 2015, Gatys et al. proposed a neural network-based style transfer method, pioneering the application of deep learning methods to image style transfer. The method of Gatys et al. has a high time complexity, and each image requires thousands of iterative optimizations. Subsequently, in order to speed up the m...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Patents(China)
IPC IPC(8): G06T3/00G06T3/40G06T5/00G06T5/40G06K9/62
CPCG06T3/4038G06T5/40G06T2207/20028G06F18/22G06T3/04G06T5/70
Inventor 马伟锋陈喆季曹婷
Owner ZHEJIANG UNIVERSITY OF SCIENCE AND TECHNOLOGY