This paper describes a method to estimate color signals in high dynamic range (HDR) scenes. Color signals of incident light into an imaging system consist of the direct spectra of light sources and the indirect spectra of the reflected lights from different object surfaces in a scene. In our study, Wiener estimation method is adopted for reconstructing color signals. The Wiener estimator requires prior statistical information such as the correlation matrix of spectral dataset and the covariance matrix of imaging noise. In Wiener estimation, the fixed matrices of imaging noise and spectral dataset are generally applied to all pixels in an image. However, the imaging noise and spectral dataset are dramatically changed in HDR scenes. Therefore, it is required to determine the suitable estimation matrix for HDR scenes. In this paper, we propose a method for determining suitable noise level and spectral dataset which are applied to Wiener estimation in HDR scenes. For validating our method, experiments using actual HDR scenes are conducted. Experimental results show the proposed method is efficient compared with the conventional Wiener estimation method, and can reconstruct accurate color signal scale in HDR scenes.