The article introduces BlazeBVD, a method for blind video deflickering that focuses on enhancing temporal consistency and eliminating flickering artifacts. The approach involves preparing flicker priors using Scale-Time Equalization (STE) in illumination space, followed by global and local flicker removal modules. The method also includes an adaptive temporal consistency model to refine video coherence. Extensive experiments demonstrate the effectiveness of BlazeBVD in achieving superior results compared to existing methods.
Sang ngôn ngữ khác
từ nội dung nguồn
arxiv.org
Thông tin chi tiết chính được chắt lọc từ
by Xinmin Qiu,C... lúc arxiv.org 03-12-2024
https://arxiv.org/pdf/2403.06243.pdfYêu cầu sâu hơn