Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper introduces a method for estimating the Structural Similarity Index (SSIM) from global Mean Square Error (MSE) for DCT-based compressed images by utilizing local error statistics derived solely from the reference image. By redistributing global MSE through variance or standard-deviation-based weighting, the authors demonstrate that their approaches yield accurate SSIM approximations that significantly outperform traditional global MSE metrics. Experiments on benchmark datasets reveal that these methods not only enhance image quality assessment but are also adaptable for video applications, making them valuable for compression and streaming systems.
SSIM can be accurately estimated from global MSE using local statistics, outperforming traditional metrics and paving the way for better image and video quality assessment.
Efficient and perceptually meaningful quality assessment is a fundamental requirement for image and video processing, compression, and streaming systems. This article shows that, in the context of Discrete Cosine Transform (DCT)-based compressed images, Structural Similarity Index (SSIM) can be approximated from global Peak Signal to Noise Ratio (PSNR) or Mean Square Error (MSE) using local statistics derived only from the reference image. While prior work assumes access to local MSE, we propose two approaches to approximate local MSE by redistributing the global MSE using variance or standard-deviation-based weighting. Experiments on the Kodak and Xiph Subset1 datasets across a range of JPEG quality levels demonstrate that both approaches provide accurate and robust SSIM approximations, substantially outperforming the global MSE baseline. The proposed framework is designed to extend naturally to video, where reference-derived statistics can be amortized across multiple encodes of the same content.