The aim of this study was to improve the spectral reflectance reconstruction precision, and to better realize the reproduction of the image color. Four different kinds of training samples were selected and the reconstruction precision of spectral reflectance of different improved Wiener estimation algorithms was compared. Four matte color cards with different color numbers such as Color Checker Color Rendition Chart, Color Checker SG, copperplate paper spot color card of Panton, and Munsell color card were chosen as training samples, and to reconstruct spectral image by the algorithms of pseudo inverse, Wiener, adaptive Wiener estimation and optimized Wiener estimation, and the results were compared. These methods of spectral reconstruction were evaluated by the root mean square error and color difference accuracy. The results showed that the reconstruction root mean square error (RMSE) based on optimized Wiener estimation was 0.00041 and the color difference was 0.7874 which were both lower than those of other kinds of algorithms. So it could be concluded that the algorithm of optimized Wiener estimation could better realize color reproduction.