A Generative Remote Sensing Image Compression Method Based on Deep Learning

A technology of remote sensing image and compression method, which is applied in neural learning methods, image communication, biological neural network models, etc.

Inactive Publication Date: 2021-03-16
WUHAN UNIV
View PDF4 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, the methods proposed by these authors are all for the compression of visible light images, not for the compression of remote sensing images.

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 Generative Remote Sensing Image Compression Method Based on Deep Learning
  • A Generative Remote Sensing Image Compression Method Based on Deep Learning
  • A Generative Remote Sensing Image Compression Method Based on Deep Learning

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0027] The specific compression process will be explained below in combination with examples and accompanying drawings.

[0028] The 3×64×64 image is used as the training image, and the 3×512×512 image is used as the test image. The main steps include:

[0029] 1. Dataset preparation and neural network hyperparameters:

[0030] 1.1 Randomly crop about 8,000 Gaofen-2 remote sensing images into image blocks with a size of 64×64×3.

[0031] 1.2 Convert the cropped image blocks into 8×64×64×3 tensors with a batch size of 8, prepare to input the network model for training, and iterate all the data 100 times. The loss function L used for training is as follows:

[0032] L=(1-MSSSIM)+MSE+0.01×PSNR+Pre_Q_diff+GAN_loss

[0033] Among them, MSSSIM represents the similarity of the multi-scale structure of the image, MSE is the mean square error, PSNR is the peak signal-to-noise ratio of the image signal (MSSSIM, MSE, and PSNR are used as the loss of the encoder network), and Pre_Q_dif ...

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 technical solution of the present invention provides a deep learning-based generative remote sensing image compression method. The present invention adopts the Pytorch deep learning framework training, takes "autoencoder (AutoEncoder) + generative confrontation model (GAN)" as paradigm, the network model is mainly divided into encoder, pre-quantization and quantization module, decoder (generator) and discrimination There are three parts in total. The present invention is suitable for compression processing of homologous remote sensing images in any spectral dimension, suitable for compression and transmission of remote sensing images under low bandwidth and low bit rate conditions, and has excellent image reconstruction capabilities. This framework is aimed at the scale and operating speed of deep neural networks It is also optimized for easy deployment and promotion for IoT devices.

Description

technical field [0001] The invention belongs to the field of remote sensing image compression, and uses a deep learning framework to compress and decompress remote sensing images. Background technique [0002] Compared with natural images, the spectral dimension of remote sensing images contains richer information, and remote sensing images have a variety of types and a large amount of data. Taking advantage of the differences in spectral curves of different ground objects, remote sensing images are widely used in various fields of the national economy. With the popularization of the application of high-resolution remote sensing imaging technology, how to effectively compress the challenges of the surge in transmission and storage data due to the significant increase in the spectral and spatial resolution of remote sensing images is an urgent problem to be solved in the application of remote sensing images. [0003] The emerging image processing method Deep Learning (Deep L...

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): H04N19/42H04N19/44H04N19/124H04N19/147G06N3/04G06N3/08
CPCH04N19/42H04N19/44H04N19/124H04N19/147G06N3/08G06N3/045
Inventor 种衍文翟亮潘少明
Owner WUHAN 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