# On the Computation of PSNR for a Set of Images or Video

arXiv: https://arxiv.org/abs/2104.14868 · HTML: https://mlsanity.com/paper/2104.14868/

## Key facts

- Cited by 13 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 30 Apr 2021.

## Built on by (most cited first, 13 of 13)

- [End-to-End Rate-Distortion Optimized Learned Hierarchical Bi-Directional Video Compression](https://mlsanity.com/paper/2112.09529/) (2021)
- [Event3DGS: Event-Based 3D Gaussian Splatting for High-Speed Robot Egomotion](https://mlsanity.com/paper/2406.02972/) (2024)
- [Generalized Event Cameras](https://mlsanity.com/paper/2407.02683/) (2024)
- [Overcoming Statistical Bias in Action-Controllable World Models](https://mlsanity.com/paper/2608.04653/) (2026)
- [Multi-Scale Deformable Alignment and Content-Adaptive Inference for Flexible-Rate Bi-Directional Video Compression](https://mlsanity.com/paper/2306.16544/) (2023)
- [STB-VMM: Swin Transformer Based Video Motion Magnification](https://mlsanity.com/paper/2302.10001/) (2023)
- [Benchmarking Video Frame Interpolation](https://mlsanity.com/paper/2403.17128/) (2024)
- [PAON: A New Neuron Model using Pad'e Approximants](https://mlsanity.com/paper/2403.11791/) (2024)
- [Explaining Image Enhancement Black-Box Methods through a Path Planning Based Algorithm](https://mlsanity.com/paper/2207.07092/) (2022)
- [FG-DFPN: Flow Guided Deformable Frame Prediction Network](https://mlsanity.com/paper/2503.11343/) (2025)
- [Pad'e Neurons for Efficient Neural Models](https://mlsanity.com/paper/2601.04005/) (2026)
- [Leveraging Image Editing Foundation Models for Data-Efficient CT Metal Artifact Reduction](https://mlsanity.com/paper/2604.05934/) (2026)
- [FashionStylist: An Expert Knowledge-enhanced Multimodal Dataset for Fashion Understanding](https://mlsanity.com/paper/2604.09249/) (2026)

## Abstract (arXiv)

When comparing learned image/video restoration and compression methods, it is common to report peak-signal to noise ratio (PSNR) results. However, there does not exist a generally agreed upon practice to compute PSNR for sets of images or video. Some authors report average of individual image/frame PSNR, which is equivalent to computing a single PSNR from the geometric mean of individual image/frame mean-square error (MSE). Others compute a single PSNR from the arithmetic mean of frame MSEs for each video. Furthermore, some compute the MSE/PSNR of Y-channel only, while others compute MSE/PSNR for RGB channels. This paper investigates different approaches to computing PSNR for sets of images, single video, and sets of video and the relation between them. We show the difference between computing the PSNR based on arithmetic vs. geometric mean of MSE depends on the distribution of MSE over the set of images or video, and that this distribution is task-dependent. In particular, these two methods yield larger differences in restoration problems, where the MSE is exponentially distributed and smaller differences in compression problems, where the MSE distribution is narrower. We hope this paper will motivate the community to clearly describe how they compute reported PSNR values to enable consistent comparison.