C.-C. Jay Kuo

TL;DR: In this article, a block-level perceptual image compression framework is proposed, including a blocklevel just noticeable difference (JND) prediction model and a preprocessing scheme, which is able to achieve 16.75% bit saving as compared to the state-of-the-art method with similar subjective quality.

TL;DR: AnomalyHop as mentioned in this paper proposes an image anomaly localization method based on the successive subspace learning (SSL) framework, which consists of three modules: 1) feature extraction via successive sub-space learning, 2) normality feature distributions modeling via Gaussian models, and 3) anomaly map generation and fusion.

TL;DR: A simple rate-distortion model for general video coders using DCT and motion compensation is developed, so that the rate and the distortion can be estimated without an expensive encoding procedure.

TL;DR: This work decomposes an input video clip into multiple smaller intervals, measure the quality of each interval separately, and applies a fusion approach to integrating these scores into a final one to improve MOVIE and is also competitive with other state-of-the-art video quality metrics.

TL;DR: A novel deep learning approach that can automatically and progressively achieve the task via adversarial blurred region mining and refining (adversarial BRMR) and outperforms the current state-of-the-art technique for blind image deblurring both quantitatively and qualitatively.