Image restoration method in extreme condition based on convolutional neural network

A technology of convolutional neural network and extreme conditions, applied in the field of image restoration, can solve the problems that cannot be widely promoted

Active Publication Date: 2019-10-25
HANGZHOU DIANZI UNIV
View PDF6 Cites 15 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0004] Thanks to the rise of neural networks in recent years, many researchers have applied neural networks to the field of image denoising, and achieved good results, but most methods are only for synthetic blurred images or only for specific noise images , and cannot be widely promoted

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 in extreme condition based on convolutional neural network
  • Image restoration method in extreme condition based on convolutional neural network
  • Image restoration method in extreme condition based on convolutional neural network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0029] The present invention will be described in detail below in conjunction with the accompanying drawings.

[0030] figure 1 It is the overall flow chart of the present invention. The system takes .RAW data as input, and its shape is H×W×1. The input data is packaged and divided into four channels. The packaged shape is H / 2×W / 2×4. The schematic diagram of the packaging process is as follows figure 2 As shown, it is a schematic diagram of the .RAW data format captured by the Sony α7S II camera. From the figure, it can be clearly seen that the R, G, and B pixels in the image array are evenly spaced, and the numbers of R, G, and B are more than It is 1:2:1. In order to make the image shape of the input neural network the same, we divide the .RAW image array into four channels, and the spatial resolution of each channel is only half of the original image.

[0031] After dividing the .RAW image array into four channels, we need to "zero" each channel, that is, subtract "Blac...

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 in an extreme condition based on a convolutional neural network. The method comprises the following steps: 1, preprocessing an acquired short-exposure image; 2, inputting the preprocessed image into a U-Net convolutional neural network for training; 3, calculating an error and carrying out iterative training; and 4, evaluating the training model,and taking the peak signal-to-noise ratio and the structural similarity of the image as evaluation criteria of a final result. According to the method, the problem of rapid imaging under low light iseffectively solved, and meanwhile, a new feasible method is provided for image denoising and deblurring.

Description

technical field [0001] The invention relates to the field of image restoration, in particular to an image restoration method under extreme conditions based on a convolutional neural network. Background technique [0002] In a low-light environment, the video images collected by the imaging system are mixed with useful information and noise due to the low ambient illumination, resulting in unclear target features, low definition, and difficult to distinguish. Digital image processing is required. Technology to reduce the impact of noise on the recognition of the scene by the human eye. At the same time, in the urban environment at night, due to the small coverage of urban lighting, there are more blind spots in urban roads. Therefore, it poses a serious challenge to nighttime video detection. The main manifestations are: first, the image presents a large number of dark areas, the content of which is blurred and the details are lost; second, under artificial light sources, s...

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): G06T5/00G06N3/04G06N3/08
CPCG06T5/002G06N3/08G06N3/045
Inventor 颜成钢王瑞海杨洪楠王兴政孙垚棋张继勇张勇东
Owner HANGZHOU DIANZI 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