Interactive Segmentation Algorithms

Abstract

In this paper, we propose a new smart interactive video object generation (SIVOG) system targeting at general semantic video object segmentation at pixel-wise accuracy. SIVOG identifies and extracts the semantic video object from image sequence with user interaction. SIVOG is a smart system that can analyze and interpret user’s high level input to define the semantic homogeneity, and adaptively process the video content. SIVOG consists of three basic components. The first one is semantic level user interaction, which allows user to enter the semantic information with the least amount of mouse clicking and movement. The second part is the smart-processing kernel, where the user-input semantic information will be analyzed and interpreted in terms of low lever features. Furthermore, the kernel adaptively determines semantic homogeneity and chooses proper algorithms to extract the semantic objects. The third component is error correction and update, where the user can correct erroneous regions in the segmentation process. The smart processing kernel can adaptively update the selected algorithms.