This paper introduces a novel feature extraction technique, named Structural Perceptual Local Feature Pattern (SPLFP). The SPLFP method integrates Fechner's law into a compact encoding scheme, aiming to mimic the human perception of external stimuli and extract features from images that conform to human visual habits. At the same time, under the framework of the famous Petersen graph, SPLFP extracts spatial structure information from four different directions within the 5×5 neighborhood of each pixel. The effectiveness of SPLFP and its single-scale methods, SPLFPv and SPLFPh, were rigorously evaluated on several face data sets. The results demonstrated that SPLFP and its single-scale methods exhibit remarkable performance, consistency, and stability across all tested datasets. This indicates that SPLFP enables the extraction of highly discriminative features, positioning it as a strong candidate for face classification tasks.