Smartphone-Based Reflectance Transformation Imaging (RTI): A Practical Relighting System with RBF/PTM Interpolation and Comparison to State-of-the-Art Techniques

In recent years, "Reflectance Transformation Imaging (RTI) has emerged as a powerful technique for capturing surface reflectance properties and enabling interactive relighting of objects". However, most RTI systems rely on fixed camera-light rigs and are not easily portable or scalable. This thesis presents a complete and practical RTI pipeline built using everyday smartphone video recordings, aimed at enabling mobile, flexible, and low-cost relighting systems. The project is structured into four main stages. First, a custom calibration module was developed to extract intrinsic camera parameters using a chessboard pattern video, enabling accurate geometric correction. Second, videos from two smartphone perspectives — a static view and a moving light source — are temporally synchronized and spatially aligned. The third stage implements Multi-Light Image Collection (MLIC) to detect markers and estimate light directions per frame, followed by interpolation using two state-of-the-art methods: Polynomial Texture Mapping (PTM) and Radial Basis Function (RBF) interpolation. These techniques allow generation of relit images under arbitrary lighting conditions. Finally, an interactive relighting application is built, allowing users to simulate dynamic light movement using mouse interaction and visualize the effect on object appearance in real time. The proposed system is modular, efficient, and implemented entirely in Python using OpenCV and NumPy, with custom modules for preprocessing, calibration, thresholding, marker detection, and interpolation. Comparative evaluation of RBF vs PTM interpolation highlights the trade-offs in accuracy and speed between the two approaches. Overall, this thesis demonstrates the feasibility of smartphone-based RTI systems and provides a functional, extensible framework for mobile reflectance analysis.