Pixel reliability assessment in optical phase unwrapping has relied predominantly on deterministicscalar metrics that cannot capture the multi-dimensional nature of phase quality in practicalinterferometric measurements. We present a multi-input fuzzy inference framework for opticalphase unwrapping that introduces linguistic reasoning for pixel reliability assessment. Unlike conventionaldeterministic approaches that rely on scalar curvature measures, the proposed frameworkevaluates pixel trustworthiness using three fuzzy variables: phase smoothness, directionalisotropy, and local consistency. This approach provides adaptive reliability assessment, a capabilitythat fixed-threshold deterministic methods fundamentally cannot achieve, without the trainingdata demands of deep learning alternatives, while maintaining exact 2π periodicity through graphbasednon-continuous propagation. Validation on simulated and experimental data demonstratesequivalent performance to conventional methods under noise-free conditions, with enhancedrobustness in challenging scenarios including moderate noise (1.67× RMS improvement) andopposing gradient discontinuities (4.4× improvement). The framework maintains sub-secondprocessing times while offering tunable parameters for application-specific optimization. Thisfuzzy reliability paradigm introduces adaptive decision-making capabilities to optical phase analysis,particularly beneficial for digital holography and interferometric applications where noise andmeasurement artifacts are prevalent.