Image restoration method based on a new encoder structure

An inpainting method and encoder technology, applied in the field of image inpainting based on a new encoder structure, can solve the problems of heavy workload and cumbersome manual inpainting, and achieve the goal of reducing the disappearance of gradients, enhancing transfer and utilization, and improving the accuracy of image inpainting. Effect

Active Publication Date: 2019-05-24
HOHAI UNIV
View PDF4 Cites 23 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, some scenes are unimaginable due to the influence of personal experience, and the workload of manual restoration is huge and tedious

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
  • Image restoration method based on a new encoder structure
  • Image restoration method based on a new encoder structure
  • Image restoration method based on a new encoder structure

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0048] In this embodiment, taking the puppy data set in ImageNet as an example, the image restoration method based on the new encoder structure of the present invention is used for image restoration, such as figure 2 As shown, the compressed feature is obtained by encoding the missing image, and the generated image is obtained by decoding the compressed feature to restore the missing image. The discriminator network discriminates the generated image and the missing image. According to the discriminative result, the generating network (encoding network and decoding network) continuously optimizes the generating result until the discriminator network cannot distinguish, that is, the best generating network is obtained through training.

[0049] It can be seen from Example 1 that the present invention can alleviate the phenomenon of gradient disappearance, enhance image feature transmission and utilization, reduce parameters, accurately extract image features without deepening ne...

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 an image restoration method based on a new encoder structureA convolutional neural network composed of an encoder and a decoder is trained to regress a missing pixel value foran image with missing pixels; The encoder captures an image context to obtain a compact feature representation, and the decoder generates missing image content using the representations. Alexnet can improve the operation speed, the network operation scale and the performance; And the Densenet can alleviate the problem of maximum gradient disappearance, enhance the feature utilization and reduce the number of parameters. According to the method, the advantages of the two are considered and combined, and a Densenet framework is added and used. Compared with an Alexnet network architecture used by an original coding and decoding device, the method has the advantages that more compact and real features can be extracted, and meanwhile, the WGAN-is used; GP adversarial loss replaces traditionalGAN adversarial loss, the feature learning speed and precision are improved, and the restoration effect is enhanced.

Description

technical field [0001] The invention belongs to image processing technology, and in particular relates to an image restoration method based on a new encoder structure. Background technique [0002] For a corrupted image, even though the central part of the image is missing, most of us can easily imagine its content based on the pixels around it, without having to see its real scene. We humans are able to understand picture structure and make visual predictions, even when only part of the scene is seen. However, some scenes are unimaginable due to the influence of personal experience, and the workload of manual restoration is huge and tedious. If you use deep learning, you can automatically fill in the missing areas of the picture, which can greatly improve the restoration efficiency. [0003] There are many existing image restoration methods, among which the image restoration method based on deep learning method is more effective. Most of the existing methods use the Alex...

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 Applications(China)
IPC IPC(8): G06T5/00
Inventor 王敏杨柳林竹
Owner HOHAI UNIV
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap