CONCOLOR: Constrained Non-Convex Low-Rank Model for Image Deblocking.

Area of Science:

• Digital Image Processing
• Signal Processing
• Computer Vision

Background:

• Block-based transform coding introduces blocking artifacts at low bitrates, limiting compression efficiency.
• Deblocking as a post-processing step offers a solution without altering existing codecs.

Purpose of the Study:

• To develop a novel image deblocking algorithm to reduce blocking artifacts and improve image quality.
• To formulate image deblocking as a constrained optimization problem within a maximum a posteriori framework.

Main Methods:

• Proposed a constrained non-convex low-rank model using an extended ℓ(p) penalty function on singular values.
• Incorporated quantization constraints into the feasible solution space.
• Developed a new quantization noise model and an alternating minimization strategy with adaptive parameter adjustment.

Main Results:

• The proposed algorithm effectively reduces blocking artifacts.
• Achieved superior performance compared to state-of-the-art methods in both objective and perceptual quality.
• The parameter-free nature makes the algorithm practical and attractive.

Conclusions:

• The novel constrained non-convex low-rank model provides an effective approach for image deblocking.
• The developed algorithm offers a practical and high-performance solution for reducing blocking artifacts in compressed images.