The light reflected from an object contains a range of information about its physical and chemical properties. Changes in the physical properties of an object can sometimes be evident as barely detectable changes of the spectra. However conventional color sensors (e.g. human color vision or typical RGB camera) are not able to detect most of spectral differences because the spectrum information is mapped into three-dimensional RGB signal. This article proposes a method for designing an optical filter which optically performs discriminant analysis of incoming spectrum to enhance object's spectral difference. First, spectral transmittance of the filter is theoretically determined by minimizing the misclassification of two selected target colors by discriminant analysis. In this study, spectral datasets of human facial colors in the presence and absence of cosmetics foundation were used as these targets. A real optical filter was developed by vacuum deposition technology according to the theoretical design, and was applied to the detection of the presence of a cosmetics foundation on the human face. In the first experiment, a digital camera equipped with the developed filter was used to detect foundation on face. The result confirmed a clear enhancement of spectral difference between facial colors with / without foundation (misclassification rate for these two sets was 4.1%), which were difficult to discriminate by naked eyes. As the second experiment, a calibration curve between the discriminant score calculated from filtered image and the amount of applied cosmetics foundation was experimentally obtained and high estimation accuracy was established (decision coefficients were about 0.9). Finally, the optical filter was applied to visualize the spatial foundation distributions of a realistic made-up skin condition. Visualized foundation map clearly indicated the unevenness of foundation distribution even though the skin color looked uniform. Results confirm that the proposed filter can achieve to visualize the foundation distribution without any destructive inspection.