Point-of-care (POC) diagnostics are increasingly shifting from centralized laboratory testing toward decentralized, rapid, and accessible platforms that can operate in resource-limited and field settings. Within this transition, computer vision (CV) has emerged as a key computational layer that converts visual diagnostic signals into quantitative, reproducible, and clinically interpretable outputs. This chapter reviews the role of CV in POC devices, with particular emphasis on smartphone-enabled imaging systems and portable optical platforms. First, the chapter outlines the broader movement from laboratory-based diagnostics to digital POC systems and introduces CV as a framework encompassing image acquisition, object detection, classification, segmentation, and image enhancement. It then examines the hardware modalities that support image-based POC analysis, including lens-based smartphone microscopy, lens-free computational imaging, and colorimetric and spectral readout systems. The chapter further discusses algorithmic frameworks ranging from classical handcrafted feature-based methods to machine learning and deep learning approaches, including recent developments in self-supervised learning and temporal modeling. Representative application areas are reviewed across hematology and cell analysis, lateral and vertical flow assays, biochemical and colorimetric assays, and infectious disease management, illustrating how CV addresses domain-specific challenges such as morphological variability, weak visual signals, and dynamic assay kinetics. Finally, the chapter highlights key barriers to clinical translation, including inter-device variability, limited annotated data, interpretability, regulatory requirements, privacy, and deployment constraints, while also discussing emerging directions such as edge computing, lens-free super-resolution imaging, and transformer-based architectures. Overall, the chapter presents CV as a central enabling layer in modern POC diagnostics, linking low-cost hardware with automated, scalable, and decentralized diagnostic interpretation.