An adaptive image compression method based on dual-map guidance

The adaptive image compression method guided by dual graphs utilizes importance and difficulty graphs to generate spatial and channel attention features, solving the problem of regional differential coding in image compression at low bit rates in existing technologies. This improves coding efficiency and reconstruction quality, making it suitable for machine vision tasks.

CN122179566APending Publication Date: 2026-06-09CHONGQING UNIV OF POSTS & TELECOMM
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
CHONGQING UNIV OF POSTS & TELECOMM
Filing Date
2026-03-05
Publication Date
2026-06-09

AI Technical Summary

Technical Problem

Existing image compression techniques struggle to achieve differentiated encoding of different regions at low bit rates, leading to blurred key structures and loss of target details, thus affecting the accuracy of machine vision tasks.

Method used

An adaptive image compression method based on dual-graph guidance is adopted. It uses a collaborative mechanism of importance graph and difficulty graph for encoding to generate spatial and channel attention-guided features. Combined with entropy coding and decoding-end gating modulation, the encoding efficiency and reconstruction quality are improved.

Benefits of technology

It achieves differentiated coding control for different regions of an image at low bit rates, improves the efficiency of bit rate resource utilization, maintains the reconstruction quality of key structures and important regions, and balances image compression performance with applicability to machine vision tasks.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122179566A_ABST
    Figure CN122179566A_ABST
Patent Text Reader

Abstract

The present application relates to the technical field of computer vision and image processing, and particularly relates to a self-adaptive image compression method based on double-map guidance, which comprises a network adopting an encoder-decoder architecture to realize compression, that is, features are extracted from an input image by an encoder, the extracted features are quantized and then subjected to entropy encoding-entropy decoding, and then a decoder maps a reconstructed image based on predicted feature values and side information obtained through entropy encoding-entropy decoding, thereby completing image compression. The present application realizes differentiated encoding control for different regions of an image, and improves the utilization efficiency of code rate resources.
Need to check novelty before this filing date? Find Prior Art