AN ADAPTIVE VLSI-BASED OBJECT TRACKING ARCHITECTURE USING LARGEST DIFFERENCE INDEXED LOCAL TERNARY PATTERNS IN THE COMPRESSED DOMAIN

Abstract

Abstract – Object tracking in dynamic and resource-constrained environments remains a critical challenge in modern computer vision systems, particularly for real-time applications such as surveillance, autonomous navigation, and smart imaging devices. This paper presents an adaptive VLSI-based object tracking architecture that leverages Largest Difference Indexed Local Ternary Patterns (LDI-LTP) in the compressed domain to achieve high efficiency and robustness. The proposed method operates directly on compressed video streams, significantly reducing computational overhead and memory requirements by avoiding full decompression.