Image rain removal method based on physical modeling generative adversarial learning

Pending Publication Date: 2021-11-23
DALIAN UNIV OF TECH
View PDF5 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

CN108765327A uses discriminative sparse coding to describe the rain line, but this will also cause color distortion to a certain extent in the output picture
CN111815526A provides a method for removing rain stripes in rainy images based on image filtering and CNN, which solves the problems that the existing rain removal methods are difficult to accurately describe the rain model, the rain stripes cannot be removed cleanly, and details are easily lost
Although the algorithm based on deep learning solves the problem of rain removal to a certain extent, it requires a large number of rain-no-rain data pairs for training to ensure the robustness of the algorithm
However, it is very difficult to simultaneously collect rain-no-rain data pairs under natural conditions, and the types of existing data are relatively single. Therefore, the current image deraining method based on deep learning is very limited in practical application. The effect is obvious for scenes similar to the characteristics of the training data, but it cannot be solved well for real scenes that are quite different from the characteristics of the training data.

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 rain removal method based on physical modeling generative adversarial learning
  • Image rain removal method based on physical modeling generative adversarial learning
  • Image rain removal method based on physical modeling generative adversarial learning

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0018] The invention is based on double-layer paradigm optimization, takes a single rainy image as input, uses the domain knowledge model and implicit prior in the lower layer problem to realize the description of the rain line in the image, and realizes the deraining to obtain a preliminary result. Then the discriminator of the upper layer problem uses the network model driven by the data to discriminate the result obtained by the lower layer, judges whether it is a clear rain-free image and returns a numerical result of 0-1, if not satisfied, the output result of the discriminator is passed through Return in the form of gradient, perform a variable update and assist the lower-level problem to continue to remove the rain. If the result is clear and there is no rain, the result of this iteration will be output as the final result obtained by the algorithm. The specific implementation plan is as follows:

[0019] The specific process of the solution network is as follows: figu...

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 field of image processing and computer vision, and provides an image rain removal method based on physical modeling generative adversarial learning, which realizes generative adversarial learning through a double-layer normal form model. The generative adversarial network is mainly composed of two parts: a generator and a discriminator, the generator generates a target result by using input random noise, and the discriminator discriminates the result generated by the generator until the discriminator considers that the result of the generator meets the target requirement. A traditional generative adversarial network is completely based on data driving, and the functions of a generator and a discriminator are achieved through a convolutional neural network. However, due to particularity of a rain removal task, a large amount of data is difficult to obtain, so that generative adversarial learning is realized from the perspective of model optimization, modeling is performed on a generator and a discriminator through an upper layer model and a lower layer model based on a double-layer normal form means, and an optimal rain removal result is obtained through the adversarial learning of the upper layer model and the lower layer model.

Description

technical field [0001] The invention belongs to the field of image processing and computer vision, and relates to an image deraining method based on physical modeling generation confrontation learning. Background technique [0002] With the development of computer hardware equipment, the processing speed of image and video data on PC and mobile terminals has been greatly improved. Life has brought many conveniences, such as: facial recognition technology when clocking in at work, reversing images when driving, traffic safety monitoring and monitoring systems, and so on. Along with it, many technologies related to image equipment and video processing are also constantly developing and improving, which in turn pushes the application range of the entire image field to become wider and wider. However, one limitation that cannot be ignored in the field of vision is the impact of severe weather. For example, in rainy weather, not only people's visual recognition of objects will ...

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
CPCG06T5/003G06N3/08G06T2207/20081G06T2207/20084G06N3/045
Inventor 刘日升姜智颖仲维樊鑫罗钟铉
Owner DALIAN UNIV 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