Image Blind Denoising System

An image and sample image technology, applied in the field of image denoising, can solve problems such as poor feasibility

Active Publication Date: 2021-06-01
HARBIN INST OF TECH
View PDF4 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0004] Aiming at the problem that existing image denoising methods rely on noise and clear image pairs for model training, and the feasibility is poor, the present invention provides a blind image denoising system

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 Blind Denoising System
  • Image Blind Denoising System
  • Image Blind Denoising System

Examples

Experimental program
Comparison scheme
Effect test

specific Embodiment approach 1

[0039] Specific implementation mode 1. Combination figure 1 and Figure 11 As shown, the present invention provides a blind image denoising system, including a self-supervised learning module and a knowledge distillation module;

[0040] The self-supervised learning modules include:

[0041] The blind spot network based on hole convolution and the image-related noise level estimation network are used to optimize the noise sample image in the noise sample image set through self-supervised loss, and obtain a preliminary blind denoising image based on Bayesian prediction, and obtain the noise sample image a first data set consisting of pairs of preliminary blind denoised images;

[0042] The image-related noise level estimation network is further used to process the clear sample images in the clear sample image set, generate corresponding noise images, and obtain a second data set composed of clear sample images and corresponding noise image pairs;

[0043] The knowledge disti...

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

An image blind denoising system belongs to the technical field of image denoising. The invention aims at the problem that the existing image denoising method relies on the pair of noise and clear image to train the model, and the feasibility is poor. Its self-supervised learning module includes: a blind spot network based on hole convolution and an image-related noise level estimation network, which are used to optimize the noise sample images in the noise sample image set through self-supervised loss, and obtain preliminary blindness removal based on Bayesian prediction The noise image is obtained to obtain the first data set; the image-related noise level estimation network also processes the clear sample image in the clear sample image set to generate a corresponding noise image, and obtains a second data set composed of the clear sample image and the corresponding noise image pair; The knowledge distillation module is used to train a multi-level wavelet-based convolutional neural denoising network using the first data set and the second data set in a fully supervised mode to obtain a denoising model. The present invention realizes blind denoising based on unpaired images.

Description

technical field [0001] The invention relates to an image blind denoising system and belongs to the technical field of image denoising. Background technique [0002] Image denoising aims to remove the noise in the image, so that the image including noise can be restored to obtain a high-quality and clear image. The main reasons for the noise in the image include the limitations of hardware conditions, the noise generated during transmission and storage, and so on. In applications such as medical images, monitoring and entertainment, higher requirements are often placed on image quality. Therefore, image denoising has always been a research hotspot in the field of image processing. [0003] In recent years, Convolutional Neural Network (CNN) has been successfully used in Gaussian denoising technology and has been continuously improved, aiming to handle more complex noise types, even real noise. However, in order to achieve a considerable denoising effect, the Gaussian denois...

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): G06T5/00G06N3/04G06N3/08
CPCG06T5/002G06N3/08G06T2207/10004G06N3/045
Inventor 武小荷刘铭曹越任冬伟左旺孟
Owner HARBIN INST OF TECH
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