Spectral reflectance reconstruction from a single RGB image of a trichromatic digital camera not only can effectively avoid the pixel distortion between the imaging channels existing in spectrum imaging system based on optical bandpass filters, but also can save the capturing time and acquire the spectral images with high spatial resolution. However, the spectral reconstruction error is influenced by reconstruction methods. A new spectral reconstruction method based on camera response expansion and local inverse distance weighted optimization is proposed. Using CIEDE2000 color difference and spectral root-mean-square error as the evaluated metrics, we use Canon 600D digital camera to verify the proposed method, and compare the proposed method with current advanced methods. The results show that the proposed method outperforms the several existing advanced methods, and average CIEDE2000 color difference and average spectral root-mean-square error of the proposed method decrease to 1.0389 and 0.0230, respectively.