Structured light depth sensing with pixelwise adaptive pattern generation for multireflective objects

Multireflective objects that exhibit spatially variant reflectance properties present a fundamental challenge to structured light three-dimensional (3D) imaging systems. Conventional methods that employ uniform illumination often result in pixel saturation in highly reflective regions and low signal-to-noise ratio in low-reflectance areas, which compromises measurement accuracy. We propose a computational imaging framework for optimal pixel-wise illumination control in depth sensing. The framework first acquires calibration images under varied illumination levels to establish a unique radiometric response curve for every camera pixel. Quadratic spline interpolation (QSI) is then employed to model these discrete response curves as continuous functions. This enables direct computation of the optimal projection intensity for each pixel to maximize fringe modulation while preventing saturation. A refined camera-to-projector coordinate mapping strategy is further introduced to address resolution disparities and resolve potential intensity conflicts in many-to-one mapping scenarios. This strategy facilitates the generation of spatially adaptive fringe patterns that maintain high signal quality across surfaces with reflectivity variations. Experimental validation on multireflective scenes demonstrates that the QSI-based method generates spatially continuous grayscale patterns for precise pixel-level adaptive projection and outperforms conventional phase-shifting profilometry and existing adaptive methods. The method is applicable to diverse industrial scenarios, including multimaterial manufacturing inspection and cultural heritage digitization.