Date of Award

12-2025

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Electrical Engineering and Computer Science

First Advisor

Marius Silaghi

Second Advisor

Ryan Stansifer

Third Advisor

Debasis Mitra

Fourth Advisor

Jian Du

Abstract

Stereo vision is a fundamental problem in computer vision, aimed at reconstructing three-dimensional scene structure from two or more two-dimensional images. Traditional stereo algorithms rely on quantitative disparity estimation, often constrained by calibration precision, lighting variations, and surface texture. In contrast, our proposed Qualitative Stereo Vision seeks to understand depth relationships and spatial configurations from multiple planar views through symbolic reasoning and constraint satisfaction, offering a more flexible and cognitively plausible approach to scene interpretation.

This dissertation presents a novel framework called Distributed Extended Waltz Filtering, designed to provide qualitative stereo vision, particularly in the presence of occlusions—a persistent challenge in multi-view interpretation. The proposed system extends the classical Waltz filtering method by distributing constraint propagation across multiple spatial regions and viewpoints. Each processing unit represents a local constraint network responsible for locally maintaining consistent labelings of visible and hidden surfaces. Communication among these distributed nodes ensures global coherence while preserving local autonomy, making the system highly scalable and fault-tolerant.

Occlusion reasoning is explicitly modeled through newly proposed qualitative constraints that capture depth discontinuities and visibility ordering among surfaces. This allows the system to infer the presence of partially hidden objects and to maintain consistent interpretations even under incomplete or ambiguous visual information. The extended framework supports iterative refinement, enabling the potentially dynamic re-evaluation of constraints as new evidence is integrated from multiple viewpoints.

A new combination operation of filtered labels is proposed and shown to satisfy logical and physical expectations and to effectively refine the sets of plausible labels when employed in Distributed Extended Waltz Filtering. Experimental results find that Distributed Extended Waltz Filtering with exploited mesh connections, besides reducing inconsistency and search space, reduces messages rates and computational overhead compared to supervisor-based qualitative filtering approaches. Furthermore, the distributed model exhibits robustness, making it suitable for applications in autonomous vision, robotic perception, and cognitive modeling.

Overall, this research contributes to the development of a distributed qualitative reasoning paradigm for stereo vision, providing a scalable and conceptually transparent method for reasoning about depth, occlusion, and spatial relations in complex visual scenes. Additionally this research introduces and compares 2 distributed algorithms for performing Distributed Waltz Filtering. We also include various case studies that explore Distributed Waltz Filtering using Multiple Agents. Lastly this research introduces new constraints that can extend the distributed Waltz Filtering Algorithm to handle occlusions which can be further explored in future works.

Download

Share

COinS