Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Single frame image rain removal method based on multi-scale feature fusion

A multi-scale feature, frame image technology, applied in the field of image processing, can solve problems such as inability to describe or detect rain, missing rain detection, and residual rain components.

Active Publication Date: 2020-10-13
SHANGHAI JIAO TONG UNIV
View PDF11 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, these traditional methods all have common problems: 1) The rainwater cannot be completely divided into high frequency in the decomposition of the image, and there will be residual rainwater components in the low frequency
2) The filter in the traditional image processing method cannot fully describe or detect all the forms of rain, which will lead to missed detection of rainwater
[0007] At present, there is no description or report of the similar technology of the present invention, and no similar data at home and abroad have been collected yet.

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
  • Single frame image rain removal method based on multi-scale feature fusion
  • Single frame image rain removal method based on multi-scale feature fusion
  • Single frame image rain removal method based on multi-scale feature fusion

Examples

Experimental program
Comparison scheme
Effect test

Embodiment

[0057] Step 1: Establish a database for training the entire deraining model.

[0058] The database includes two parts: training set and test set. The training set contains rainy images and their corresponding non-rainy images. In this embodiment, the training set includes 700 pictures, 500 of which are selected from the first 800 pictures in the UCID image database, and the remaining 200 are from the training set in the BSD-500 database. Among them, the rainy image is synthesized by artificially adding rain lines, and the shape, intensity and direction of the rain lines are diverse to ensure the generalization ability. The test set is divided into two categories: the synthetic picture test set and the natural picture test set; the synthetic picture test set contains the original image without rain, so the peak signal-to-noise ratio (PSNR) on the luminance channel (luminance channel) can be calculated at last Equal quantitative indicators to evaluate the image quality (the br...

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 present invention provides a single-frame image deraining method based on multi-scale feature fusion, which extracts features of rainy images by using receptive fields of different scales, and then obtains deraining results through features through deconvolution operations. The combination of scale features and fine-scale features promotes the rain-free images generated at fine scales to achieve the best effect of removing rain. By removing rainlines on multiple scales, it can be used in a variety of rainwater situations, and the rain removal algorithm is more universal. The present invention constructs a new error function by citing the confrontation error and the perception error, and trains the deraining model without any prior knowledge or pre-processing and post-processing of the image, thus ensuring the integrity of the entire structure; The results on multiple test sets show that the present invention can improve the peak signal-to-noise ratio on the luminance signal channel by 2-5dB compared with the result of the leading mainstream rain removal algorithm.

Description

technical field [0001] The present invention relates to the technical field of image processing, in particular, to a single-frame image rain removal method based on multi-scale feature fusion, which is applied to processing a single-frame image with rain, so as to remove the rain in the image, restore the background image and restore the color of the original image. and detail to minimize distortion. Background technique [0002] In complex and changeable weather conditions, the pictures or videos taken by people are often disturbed or blurred by rain, snow, or fog, or the subject of the picture will be seriously disturbed. Rain, as one of the most common natural phenomena in life, will degrade human vision. In the case of rain, the distance rain line will have a reflection effect under the influence of light, and the quality of the captured pictures and videos will be seriously degraded. At the same time, the rain line will accumulate to form fog, and there will be rain an...

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
CPCG06N3/084G06N3/048G06T5/73
Inventor 徐奕张峥倪冰冰杨小康
Owner SHANGHAI JIAO TONG UNIV
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products