Segment-based depth maps extraction from stereo video content

摘要

In this paper a content-independent, adaptive and parameter-free estimation chain for depth map generation from still-images stereo content will be described. The state-of-the-art solutions for each step of the classical chain will be analyzed as reference benchmark for our proposed implementation. Finally, comparative results will be presented and evaluated through the Middlebury Stereo Evaluation framework to prove the achieved optimal trade-off between quality and computational complexity.