Method for removing blocking effect of JPEG image based on relevant structure group pareto distribution model

By constructing a related structure set and a generalized Pareto distribution model, and combining it with Bayesian criterion optimization, the problem of insufficient utilization of nonlocal similarity in existing JPEG image deblocking methods is solved, and a more efficient image deblocking effect is achieved.

CN122243811APending Publication Date: 2026-06-19CHONGQING UNIV
View PDF -1 Cites -1 Cited by

Patent Information

Authority / Receiving Office
CN Β· China
Patent Type
Applications(China)
Current Assignee / Owner
CHONGQING UNIV
Filing Date
2026-04-20
Publication Date
2026-06-19

AI Technical Summary

Technical Problem

Existing JPEG image deblocking methods struggle to effectively utilize the nonlocal similarity and diverse prior information of images, leading to a decline in image quality after low bit rate compression.

Method used

A related structure group is constructed, and sparse representation is performed using a PCA dictionary. It is assumed that the coefficients follow a generalized Pareto distribution. A unified model for coefficient estimation and quantization noise removal is established by combining the Bayesian criterion. The block artifact removal and image edge texture detail restoration of JPEG images are achieved through optimization solution.

Benefits of technology

It significantly improves the performance of image deblocking by accurately constructing relevant structure groups and comprehensively considering the correlation between structure groups, thereby enhancing the structured sparsity of coefficients and improving image quality.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122243811A_ABST
    Figure CN122243811A_ABST
Patent Text Reader

Abstract

This invention discloses a JPEG image deblocking method based on a Pareto distribution model of a correlation structure set. It belongs to the field of digital image processing technology. First, it constructs a correlation structure set and learns a PCA dictionary to sparsely represent the correlation structure set. Based on the assumption that the coefficients follow a generalized Pareto distribution, it establishes a unified model for coefficient estimation and quantization noise deblocking under a Bayesian criterion. Finally, it optimizes the solution model to remove block artifacts and restore image edge texture details in JPEG images. This invention uses the correlation structure set as the processing object, fully exploring the correlation between coefficients of different structure sets, and uses the generalized Pareto probability distribution for coefficient estimation. Furthermore, it employs an alternating direction iterative algorithm to efficiently solve sub-problems related to structure set coefficient estimation and image reconstruction. This invention improves the estimation accuracy of JPEG image coefficients, resulting in richer image details after deblocking, and therefore can be used for JPEG image deblocking.
Need to check novelty before this filing date? Find Prior Art