Silicon carbide (SiC) is a critical wide-bandgap semiconductor widely used in high-power and high-temperature applications. However, microscopic defects in SiC wafers can significantly degrade device performance, and conventional methods for defect inspection are often expensive, bulky, and unsuitable for inline industrial deployment. In this study, we propose a cost-effective hyperspectral defect detection framework that reconstructs spectral reflectance from conventional RGB images and applies spectral similarity analysis to identify defects in SiC materials. A color calibration target with known spectral reflectance was first used to establish a linear RGB-to-spectral transformation matrix. By averaging RGB responses over individual color patches, a stable camera response was obtained and mapped to 61 spectral bands (400–700 nm at 5 nm intervals) using a pseudoinverse-based calibration approach. The reconstructed spectra of test images were then analyzed using spectral angle mapper (SAM) and CIE ΔE color difference metrics. SAM captures spectral shape variations, while ΔE reflects perceptual energy differences, enabling complementary defect characterization. A robust statistical thresholding method combined with morphological filtering was applied to generate defect maps without requiring supervised labels. Experimental results demonstrate that the proposed method successfully highlights subtle surface defects in SiC samples and produces spatially consistent defect localization results comparable to those obtained from true hyperspectral systems. This work shows that RGB-based spectral reconstruction, combined with physically meaningful spectral metrics, provides a practical and low-cost alternative for hyperspectral defect inspection in semiconductor manufacturing.