Video Sky Replacement Based on Dual-Branch Segmentation Network
Xiangrong Fu, Binzi Wang, Yang Shu, Wanyu Yang, Liya Tian, Youdong Ding · 2024
This article presents a video sky replacement technology. The core method of this technology is a semantic detail dual-branch sky segmentation network. It accurately extracts the sky area in the video using advanced segmentation technology. Subsequently, feature points in the sky area are extracted to calculate optical flow, enabling the tracking and identification of dynamic elements in the sky. By adjusting the image filter parameters, the color and brightness of the composite image can be accurately adjusted to seamlessly blend the replaced sky with the original video. This technology simplifies the video post-production process and provides creators with efficient tools to transform the world with just one click. It opens up new possibilities for enhancing visual effects and creative expression.