Skip to content

Official implementation of "Early Exit or Not: Resource-Efficient Blind Quality Enhancement for Compressed Images" (ECCV'20).

License

Notifications You must be signed in to change notification settings

ryanxingql/rbqe

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

46 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Early Exit or Not: Resource-Efficient Blind Quality Enhancement for Compressed Images (ECCV 2020)

Background

Official repository of Early Exit or Not: Resource-Efficient Blind Quality Enhancement for Compressed Images, ECCV 2020. [速览]

  • A single blind enhancement model for HEVC/JPEG-compressed images with a wide range of Quantization Parameters (QPs) or Quality Factors (QFs).
  • A multi-output dynamic network with early-exit mechanism for easy input.
  • A Tchebichef-moments based NR-IQA approach for early-exit decision. This IQA approach is highly interpretable and sensitive to blocking energy detection.

network

Feel free to contact: [email protected].

Code & Pre-trained Model

We have released two versions of the RBQE approach.

  • ECCV paper version

    • Adopts the RAISE data-set (high-quality and large-scale).
    • Adopts the HM software for compression and get YUV images.
    • Enhances only Y channel and report the Y-PSNR result (in accordance to previous papers).
    • Implements the image quality assessment module with MATLAB (convenient for visualization).
    • Assesses the Y quality; the IQA threshold is determined according to the Y performance.
  • Improved version

    • Adopts the DIV2K data-set (used by most image restoration approaches).
    • Adopts the BPG software for compression (faster, but the result is different to that of the HM) and get PNG images (simpler).
    • Enhances RGB channels (more practical).
    • Re-implements the MATLAB-based image quality assessment module with Python (no need to use MATLAB anymore; but much, much slower than the MATLAB version).
    • Assesses the R quality; the IQA threshold is determined according to the R performance.

We have released the codes of all compared approaches in the latter repository.

About

Official implementation of "Early Exit or Not: Resource-Efficient Blind Quality Enhancement for Compressed Images" (ECCV'20).

Resources

License

Stars

Watchers

Forks

Languages