Detachable Object Detection: Segmentation and Depth Ordering From Short-Baseline Video

TL;DR

This paper presents an unsupervised method for segmenting and detecting detachable objects in short-baseline videos by integrating appearance and motion cues into a cost functional optimized via linear programming.

Contribution

It introduces a novel unsupervised approach that combines appearance and motion data for segmenting and depth ordering of objects in short video sequences.

Findings

Effective segmentation of scene surfaces using minimal video data

Capability to seed longer-term optimization for improved results

Demonstrates potential and limitations of the approach

Abstract

We describe an approach for segmenting an image into regions that correspond to surfaces in the scene that are partially surrounded by the medium. It integrates both appearance and motion statistics into a cost functional, that is seeded with occluded regions and minimized efficiently by solving a linear programming problem. Where a short observation time is insufficient to determine whether the object is detachable, the results of the minimization can be used to seed a more costly optimization based on a longer sequence of video data. The result is an entirely unsupervised scheme to detect and segment an arbitrary and unknown number of objects. We test our scheme to highlight the potential, as well as limitations, of our approach.