Real-Time Object-Based Video Segmentation Using Colour Segmentation and Connected Component Labeling

摘要

In this paper, we described two-scan connected component labeling (CCL) approach on a real-time colour video image segmentation. CCL approach is an act of region labeling and could provides opportunity to find feature of object and establish boundaries of objects which are the common properties needed by many object-based video segmentation applications. We tested the proposed technique in two experimental studies that simulates real-time object-based video segmentation. Our experiments results shown that the proposed technique could perform region labeling in a fast manner. Another advantage of the proposed technique is that it does not provide extra storage to store same label equivalence. This property gives advantage to avoid label equivalence redundancies that always happen in the CCL approach.