Image compression method and system based on octave convolution and semantic segmentation

A semantic segmentation and image compression technology, applied in the field of computer vision, can solve problems such as poor image compression performance, artifacts, high patent fees, etc., and achieve the effects of saving computing resources, ensuring compression rate, and saving space

Active Publication Date: 2021-05-04
SHANDONG NORMAL UNIV
View PDF2 Cites 2 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

BPG uses HEVC as the encoding core. Under the same resolution, the BPG file size is half that of JPEG. Although BPG is very good, the high patent fee prevents it from being widely used.
These traditional image compression schemes rely on the individual optimization of the encoder, the processing flow is relatively complex, sometimes unsatisfactory in terms of detail restoration, usually accompanied by artifacts, blurring and other shortcomings, and the performance of low data rate image compression is also poor. poor

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 compression method and system based on octave convolution and semantic segmentation
  • Image compression method and system based on octave convolution and semantic segmentation
  • Image compression method and system based on octave convolution and semantic segmentation

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0038] The purpose of this embodiment is to provide an image compression method based on octave convolution and semantic segmentation.

[0039] An image compression method based on octave convolution and semantic segmentation, including:

[0040] Use the pre-trained semantic segmentation network to generate a semantic segmentation map of the original image and encode it losslessly;

[0041] Using the semantic segmentation map and the original image as the input of the first set of octave convolutional networks, generating a compressed representation of the image, and encoding it losslessly; upsampling the compressed representation, and combining it with the semantic The segmentation map is used as the input of the second group of octave convolutional networks to obtain the estimated value of the original image;

[0042] Calculate the residual between the original image and the estimated value of the original image, perform lossy coding on the residual, add the estimated value...

Embodiment 2

[0082] The purpose of this embodiment is to provide an image compression system based on octave convolution and semantic segmentation.

[0083] An image compression system based on octave convolution and semantic segmentation, including:

[0084] A semantic segmentation map acquisition unit, which is used to generate a semantic segmentation map of the original image using a pre-trained semantic segmentation network, and perform lossless encoding on it;

[0085] A coding unit, which is used to use the semantic segmentation map and the original image as the input of the first group of octave convolutional networks, generate a compressed representation of the image, and perform lossless encoding on it; upsample the compressed representation, and Using it and the semantic segmentation map as the input of the second group of octave convolutional networks to obtain the original image estimation value; meanwhile, calculate the residual error between the original image and the origina...

Embodiment 3

[0088] The purpose of this embodiment is to provide an electronic device, including a memory, a processor, and computer instructions stored in the memory and run on the processor. When the computer instructions are run by the processor, the above-mentioned Octave-based Image compression methods for convolution and semantic segmentation, including:

[0089] Use the pre-trained semantic segmentation network to generate a semantic segmentation map of the original image and encode it losslessly;

[0090] Using the semantic segmentation map and the original image as the input of the first set of octave convolutional networks, generating a compressed representation of the image, and encoding it losslessly; upsampling the compressed representation, and combining it with the semantic The segmentation map is used as the input of the second group of octave convolutional networks to obtain the estimated value of the original image;

[0091] Calculate the residual between the original im...

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 provides an image compression method and system based on octave convolution and semantic segmentation. The scheme comprises the following steps: generating a semantic segmentation map of an original image by using a pre-trained semantic segmentation network; taking the semantic segmentation map and the original image as input of a first group of octave convolution networks, and generating a compressed representation of the image; performing up-sampling on the compressed representation, and taking the compressed representation and the semantic segmentation map as input of a second group of octave convolution networks to obtain an original image estimation value; and calculating a residual error between an original image and the original image estimation value, carrying out lossy coding on the residual error, and adding the original image estimation value and a decoded residual error image to obtain a final reconstructed image. According to the method, the distribution of bit streams in the image space can be guided by inputting the semantic segmentation map, high-frequency information and low-frequency information in the image space are independently optimized by using the octave convolution network, and a high-quality and high-detail image can be effectively synthesized while the compression ratio is ensured.

Description

technical field [0001] The present disclosure relates to the technical field of computer vision, in particular to an image compression method and system based on octave convolution and semantic segmentation. Background technique [0002] The statements in this section merely mention background art related to the present disclosure and do not necessarily constitute prior art. [0003] With the rapid development of information technology, more and more multimedia data appear on the Internet, and high-quality multimedia content begins to be widely popularized. How to process these data to improve transmission efficiency and reduce storage cost has become an extremely important issue. For images, a large number of images generate a huge amount of data, which brings great challenges to storage and transmission. Reasonable image compression is conducive to reducing storage pressure and improving transmission capabilities. Image compression technology has received more and more at...

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): G06T9/00G06N3/04
CPCG06T9/002G06N3/04
Inventor 孟丽丽刘志远蔡晓雅张佳谭艳艳张化祥
Owner SHANDONG NORMAL 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