A multi-scale face image deblurring algorithm based on separation of low frequency and high frequency

A face image, multi-scale technology, applied in image enhancement, image data processing, computing and other directions, can solve the problems of limited help of face deblurring algorithms, insufficient use of features, and insufficient feature extraction, and achieve rich details. , Enhanced effect, structure and detail complete effect

Active Publication Date: 2019-06-18
FUDAN UNIV
View PDF5 Cites 16 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Although these methods have tried to use multi-scale features, their use of features is not comprehensive enough, they only scale the image, and the feature extraction is not sufficient.
[5] Use the semantic segmenta

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
  • A multi-scale face image deblurring algorithm based on separation of low frequency and high frequency
  • A multi-scale face image deblurring algorithm based on separation of low frequency and high frequency
  • A multi-scale face image deblurring algorithm based on separation of low frequency and high frequency

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0040] For a blurred face picture, to deblur it, you can use figure 1 The method described is implemented.

[0041] The specific process is:

[0042] 1. Training

[0043] (1) Use the pre-processing network to simultaneously generate low-frequency information and semantic segmentation of pictures

[0044] Simultaneously fit low-frequency information and semantic segmentation using a pre-processing network. The weighted pre-training pre-processing network using semantic segmentation and low-frequency information loss function;

[0045] (2) Using multi-scale networks to restore clear pictures

[0046] The semantic segmentation, low-frequency information, and blurred images generated by the pre-processing network are input into the multi-scale network for deblurring. The loss function is multi-scale content loss;

[0047] (3) Joint high-level task training

[0048] The pre-processing network and the multi-scale network are collectively referred to as the deblurring network. ...

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 belongs to the technical field of digital image intelligent processing, and particularly relates to a multi-scale face image deblurring algorithm based on separation of low frequency andhigh frequency. The algorithm comprises the steps that a preprocessing network is used, and low-frequency information and semantic segmentation of a face image are recovered at the same time; Then the generated low-frequency information, semantic segmentation and fuzzy picture splicing are input into a subsequent multi-scale network to recover a clear picture; And finally, the deblurring network(a pre-processing network and a multi-scale network) and a high-level task are jointly trained, so that a clear picture generated by deblurring can be better expressed on the recognition of equal-level tasks. Experimental results show that the restored clear picture is very complete in structure and details, and meanwhile, the restored clear picture contains richer perceptual information through combination with a high-level task, so that the effect of the high-level task is greatly improved.

Description

technical field [0001] The invention belongs to the technical field of digital image intelligent processing, and specifically relates to a face image deblurring algorithm, more specifically, relates to a multi-scale face image deblurring algorithm based on separating low and high frequencies. Background technique [0002] In recent years, with the popularity of mobile devices, taking pictures has become an indispensable part of daily life. However, holding mobile devices such as mobile phones will inevitably cause shaking when taking pictures, which will lead to motion blur in the image, especially in low-light scenes, which require a longer exposure time. Motion blur can have a severe impact on high-level tasks. Taking the face image as an example, blurring the face will cause difficulties in face recognition and face key point detection. Face image deblurring, which can deblur the blurred face image and restore the structure and details of the face. It plays an importan...

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): G06T5/00G06K9/00
Inventor 颜波李昂
Owner FUDAN 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