Eureka AIR delivers breakthrough ideas for toughest innovation challenges, trusted by R&D personnel around the world.

Image restoration system based on multilevel wavelet convolutional neural network

A convolutional neural network and convolutional neural technology, applied in the field of image restoration systems, can solve problems such as the inability to take into account image restoration quality and image restoration speed

Active Publication Date: 2018-11-06
HARBIN INST OF TECH
View PDF4 Cites 6 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0004] In order to solve the problem that the existing image restoration system based on convolutional neural network cannot take into account both image restoration quality and image restoration speed, the present invention proposes an image restoration system based on multi-level wavelet convolutional neural network

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 system based on multilevel wavelet convolutional neural network
  • Image restoration system based on multilevel wavelet convolutional neural network
  • Image restoration system based on multilevel wavelet convolutional neural network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment

[0090] Embodiment: combine below Figure 1 to Figure 7 This embodiment will be described in detail.

[0091] refer to figure 1 , the image restoration system based on the multi-level wavelet convolutional neural network described in this embodiment includes a contraction sub-network and an expansion sub-network;

[0092] The contraction subnetwork includes an input encoder and a wavelet transformer. The input encoder includes M convolutional neural subnetworks. The wavelet transformer includes M wavelet transform layers. The wavelet transform layers and convolutional neural subnetworks are arranged alternately. The output of the former connected to the input of the latter;

[0093] The extended subnetwork includes an output decoder and a wavelet inverse transformer, the output decoder includes M deconvolutional neural subnetworks, the wavelet inverse transformer includes M wavelet inverse transform layers, and the deconvolutional neural subnetwork and wavelet inverse transfo...

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 system based on a multilevel wavelet convolutional neural network, belongs to the field of image restoration, and solves the problem that an existing imagerestoration system based on a convolutional neural network cannot give consideration to image restoration quality and image restoration speed. According to the system, wavelet transform layers and convolutional neural sub-networks are arranged alternately; the output ends of the wavelet transform layers are connected with the input ends of the convolutional neural sub-networks; the input end of afirst deconvolutional neural sub-network is connected with the output end of the Mth convolutional neural sub-network; the output end of the first wavelet transform layer is further connected with the input end of an Mth invert wavelet transform layer; the output ends of the first to (M-1)th convolutional neural sub-networks are connected with the input ends of the Mth to second deconvolutional neural sub-networks; an input object of the first wavelet transform layer is a to-be-restored image; and an output result of the Mth invert wavelet transform layer is a restored image; and M is greaterthan or equal to 2.

Description

technical field [0001] The invention relates to an image restoration system, which belongs to the field of image restoration. Background technique [0002] In real life, due to the limitation of the image acquisition resolution of the imaging system hardware, the movement of the camera and the change of the light environment during the image acquisition process, the limitation of storage space and transmission bandwidth, the image often needs to be down-sampled and lossy compressed. , resulting in low-quality images such as blurring, noise interference, and low resolution. Such low-quality images not only affect people's visual senses, but also cannot be applied to artificial intelligence applications such as face recognition and object detection that require high-quality input images. Therefore, it is particularly necessary to restore low-quality images to high-quality images. [0003] In recent years, scholars have attempted to apply convolutional neural networks to the ...

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): G06T5/00G06N3/04
CPCG06T2207/20064G06N3/045G06T5/00
Inventor 左旺孟张宏志刘鹏举
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
Eureka Blog
Learn More
PatSnap group products