Image noise reduction method utilizing depth full-convolution coding-decoding network

A technology of image noise reduction and convolutional coding, applied in the field of image noise reduction, to achieve the effect of benefiting the network level, optimizing noise reduction performance, and eliminating noise

Inactive Publication Date: 2016-12-21
TONGJI UNIV
View PDF0 Cites 33 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0002] Image denoising is a classic problem in the field of low-level vision, it has been widely s

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 noise reduction method utilizing depth full-convolution coding-decoding network
  • Image noise reduction method utilizing depth full-convolution coding-decoding network
  • Image noise reduction method utilizing depth full-convolution coding-decoding network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment

[0034] An image denoising method using a deep fully convolutional encoder-decoder network. The input of the model is a noisy image and the output is a clean image after denoising. The model has a very deep encoder-decoder framework. Different from traditional approaches that use prior knowledge of images, this framework directly learns from the training dataset an end-to-end fully convolutional mapping that maps input noisy images to clean output images. The network model consists of multiple layers of convolution and deconvolution operations. Since the deeper the network training is more difficult, this method uses the inter-layer skip connection structure to connect the convolutional layer and the symmetrical deconvolutional layer, so that the training converges faster and a higher quality local optimum can be obtained untie.

[0035] Model framework such as figure 1 As shown, it mainly includes a series of convolutional layers and a symmetrical deconvolutional layer. Th...

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 relates to an image noise reduction method utilizing a depth full-convolution coding-decoding network. According to the method, a neural network model is employed to carry out image noise reduction, input of the model is an image with noise, and output of the model is the image after noise reduction. The method is characterized in that the model employs a symmetric coding-decoding network structure and comprises N convolution layers and N de-convolution layers which are sequentially connected, an ith convolution layer and an (N+1-i)th de-convolution layer are in symmetric relationship, i=1, 2, and so on, and N, the convolution layer is taken as a characteristic extractor and is used for coding main content of the image and carrying out noise reduction, and the de-convolution layer is used for decoding abstract content of the image and recovering detailed content of the image. Compared with the prior art, noise reduction is carried out while the main content of the image is coded, moreover, the de-convolution layer is used for decoding the abstract content of the image and recovering the detailed content of the image, and the detailed content of the image can be kept to the maximum degree.

Description

technical field [0001] The invention relates to an image noise reduction method, in particular to an image noise reduction method using a deep full convolutional encoding-decoding network. Background technique [0002] Image denoising is a classic problem in the field of low-level vision, and it has been widely studied in the fields of image processing and computer vision, but it is still a challenging problem. Traditional image denoising methods are mainly based on wavelet shrinkage, full difference or prior knowledge of the image. Among them, the method based on wavelet shrinkage models the wavelet transform coefficients as a Laplace distribution model. The method based on full difference assumes that the gradient of the image conforms to the Laplacian distribution. One representative of the methods based on image prior knowledge is the dictionary-based method. Benefiting from the success of the K-SVD algorithm, methods for learning dictionaries for image denoising have...

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
IPC IPC(8): G06T5/00G06N3/08
CPCG06N3/084G06T5/002G06T2207/20182
Inventor 尤鸣宇沈春华王慧慧
Owner TONGJI 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