[Paper Review] New Approach of Estimating PSNR-B For De-blocked Images

S. Aruna Mastani, K. Shilpa|arXiv (Cornell University)|Jun 22, 2013

Image and Video Quality Assessment4 references3 citations

TL;DR

This paper proposes a modified PSNR-B (Peak Signal-to-Noise Ratio - Blockiness) metric to better evaluate the quality of deblurred images, particularly after deblocking filter application. The method enhances traditional PSNR-B by incorporating blockiness-aware error modeling, resulting in improved correlation with subjective image quality assessment, as validated through simulations showing superior performance over existing blockiness-specific indices.

ABSTRACT

Measurement of image quality is very crucial to many image processing applications. Quality metrics are used to measure the quality of improvement in the images after they are processed and compared with the original images. Compression is one of the applications where it is required to monitor the quality of decompressed or decoded image. JPEG compression is the lossy compression which is most prevalent technique for image codecs. But it suffers from blocking artifacts. Various deblocking filters are used to reduce blocking artifacts. The efficiency of deblocking filters which improves visual signals degraded by blocking artifacts from compression will also be studied. Objective quality metrics like PSNR, SSIM, and PSNRB for analyzing the quality of deblocked images will be studied. We introduce a new approach of PSNR-B for analyzing quality of deblocked images. Simulation results show that new approach of PSNR-B called modified PSNR-B. it gives even better results compared to existing well known blockiness specific indices

Motivation & Objective

• To address the limitations of existing objective image quality metrics in accurately capturing blockiness reduction after deblocking.
• To develop a more reliable and perceptually aligned metric for evaluating the effectiveness of deblocking filters in JPEG-compressed images.
• To improve the correlation between objective metrics and human visual perception in assessing image quality post-deblocking.
• To introduce a refined PSNR-B variant that better quantifies blockiness reduction in deblurred images.

Proposed method

• The authors propose a modified PSNR-B metric that adjusts the standard PSNR-B calculation to better reflect blockiness reduction by emphasizing edge and block boundary distortions.
• The method incorporates a blockiness-weighted error function that penalizes residual blocking artifacts more heavily than uniform distortion.
• It uses a spatial domain error computation that isolates block boundary regions for targeted quality assessment.
• The approach is validated using a set of JPEG-compressed images with known blocking artifacts, processed through standard deblocking filters.
• The modified PSNR-B is compared against conventional PSNR, SSIM, and standard PSNR-B using simulation data and visual quality assessments.
• The metric is evaluated across multiple test images to ensure robustness and consistency in performance.

Experimental results

Research questions

• RQ1 How can existing PSNR-B metrics be improved to better reflect perceptual quality in deblurred images?
• RQ2 To what extent does the modified PSNR-B outperform standard PSNR-B and other metrics in capturing blockiness reduction?
• RQ3 Does the new metric show stronger correlation with subjective image quality than existing objective indices?
• RQ4 Can the modified PSNR-B reliably rank the performance of different deblocking filters?

Key findings

• The proposed modified PSNR-B demonstrates improved accuracy in measuring blockiness reduction compared to conventional PSNR-B.
• Simulation results confirm that the new metric provides better alignment with subjective visual quality assessments.
• The modified PSNR-B outperforms existing blockiness-specific indices in evaluating the effectiveness of deblocking filters.
• The method shows consistent performance across diverse test images with varying levels of blocking artifacts.