Content-adaptive video compression doesn’t miss a thing

Опубликовано 17 февраля 2025, 12:34

Neural networks that are trained too closely on big datasets may struggle to reconstruct compressed video accurately. Content-adaptive neural network tools such as loop-filters can positively exploit this overfitting, improving coding efficiency with negligible decoding costs. Find out how content-adaptation can also enable decoding customization for specific content in addition to these major coding gains.

Discover the leading research by Ruiying Yang, Maria Santamaria, Francesco Cricri, Honglei Zhang, Jani Lainema, Ramin G. Youvalari, Miska M. Hannuksela and Tapio Elomaa.

ieeexplore.ieee.org/document/10402710