Unlock instant, AI-driven research and patent intelligence for your innovation.

A UAV Image Denoising Method Based on Fully Convolutional Siamese Network

A twin network and unmanned aerial vehicle technology, applied in the field of image processing, can solve problems such as ignoring the image block structure and inaccurate similar block groups

Active Publication Date: 2022-07-05
FUZHOU UNIV
View PDF5 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0003] In view of this, the object of the present invention is to provide a method for denoising UAV images based on a fully convolutional twin network, which solves the problem that the block matching algorithm is not accurate enough and ignores the image block structure when using Euclidean distance to find similar block groups

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 UAV Image Denoising Method Based on Fully Convolutional Siamese Network
  • A UAV Image Denoising Method Based on Fully Convolutional Siamese Network
  • A UAV Image Denoising Method Based on Fully Convolutional Siamese Network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0059] The specific steps of applying denoising to real images in this embodiment are as follows:

[0060] 1. On the clean image dataset {y (1) ,y (2) ,...,y (m) } Add Gaussian noise with mean 0 and standard deviation σ=5-10 randomly three times to get the training noise image set {x (1) ,x (2) ,...,x (m) }, where m=45;

[0061] 2. With the help of {x (1) ,...,x (m) } and {y (1) ,...,y (m) } Calculate Mahalanobis distance to get similar block labels Among them, the number of positive sample labels M=10, and the number of negative samples is 4 times the number of positive samples;

[0062] 3. Put {x (1) ,...,x (m) } Input into the neural network f to get the output feature map

[0063] 4. Pass get the corresponding channel vector of the label

[0064] 5. Minimize the objective function to optimize the network: β=100;

[0065] 6. Repeat steps 2)-5) until the number of iterations requirements are met.

[0066] 7. Input the image x to be denoised into the ...

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 relates to a denoising method for unmanned aerial vehicle images based on a fully convolutional twin network. The Mahalanobis distance is used to calculate similar block group labels in a combination of clean and noisy images; then a fully convolution twin network is established for training. In the denoising scene, the Siamese network outputs similar block groups of the image to be denoised, and then constructs a hybrid orthogonal dictionary with external and internal information through a Gaussian mixture model. Finally, the weighted sparse coding framework is used to solve the orthogonal dictionary to reconstruct the image blocks, and then the image blocks are aggregated to achieve the final denoising. The invention uses Mahalanobis distance and full convolution twin network to find similar block groups, and solves the problem that the block matching algorithm is not accurate enough and ignores the image block structure when using Euclidean distance to find similar block groups.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a method for denoising an unmanned aerial vehicle image based on a fully convolutional twin network. Background technique [0002] UAVs have the characteristics of strong mobility and high efficiency, and are currently used in a wide range of fields such as aerial photography, agriculture, disaster rescue, and power inspection. UAVs are equipped with high-precision cameras to obtain image data on the ground, thereby helping people to accurately analyze information in real time. However, due to the vibration of the body of the drone during the shooting process, the captured images often contain noise, so how to remove the noise to obtain accurate information of the real image has become a difficulty. Image denoising refers to recovering clean images from noise-contaminated images, and is the basis for improving the performance of advanced computer vision tasks such as cl...

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): G06T5/00G06K9/62G06N3/04G06N3/08G06V10/74G06V10/82
CPCG06N3/04G06N3/08G06F18/22G06T5/70Y02T10/40
Inventor 陈飞尤福源
Owner FUZHOU UNIV