Â鶹¾«Æ·ÊÓƵ

Eimran Eimon recently presented his research at the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) 2026 in Denver, Colorado. Widely regarded as the world's leading conference in computer vision and artificial intelligence. His work is inspired by how humans perceive images while judging quality, the proposed model leverages most and least apparent distortions that are analogous to global and local appearance in humans.


At the conference, Eimon presented his paper titled "PSIM: Perceptual Similarity Index Measure," which introduces a new approach to measuring image quality. The proposed metric is based on Multi-level Wasserstein Distortion (MWD) and was shown to outperform widely used image quality measures such as PSNR, SSIM, and LPIPS. The results establish PSIM as a new state-of-the-art method for evaluating perceptual image quality.

One of the paper's key contributions is the introduction and formalization of two important concepts in image perception: Most Apparent Distortion (MAD) for global deviations that are immediately apparent and Least Apparent Distortion (LAD) for local deviations which are evident upon closer inspection. Through extensive experiments, this research demonstrates how these two types of distortions jointly work together to influence the way people perceive image quality.

Read the full paper here:

Md Eimran Hossain Eimon and Hari Kalva, PSIM: Perceptual Similarity Index Measure . CVPR 2026, pp. 8564–8574.

Paper link: