How Mux detects shot boundaries
Blog post from Mux
Shots is a new feature from Mux Video that creates a manifest of shot boundaries and representative images for video content, forming the basis of the Mux Robots workflow, Find Scenes. The algorithm behind Shots is designed to detect shot boundaries efficiently using minimal resources, prioritizing false positives over false negatives. It operates by analyzing pixel changes between frames, utilizing metrics such as the sum of absolute differences and the mean average frame difference (MAFD) from FFmpeg. Initial approaches involved setting score thresholds and searching for peaks, each with their unique limitations, leading to false positives. The final solution combines both approaches, detecting shot changes when both methods agree, using the last three values for a resource-efficient, one-pass process. While effective, the algorithm struggles with smooth transitions, which can result in false negatives, and with videos containing flashing lights, which can lead to false positives.
No tracked trend matches for this post yet.