Multispectral reconstruction technology has widespread potential applications in biomedicine. Moreover, visual inspection is the typically used traditional method for skin and oral bacteria assessment. However, this method relies on naked-eye observations and is highly subjective. Even experienced clinicians face uncertainty while providing a definite diagnosis for the bacterial infection. This study proposes a fluorescence multispectral Wiener estimation-based reconstruction method to provide an objective and accurate bacterial evaluation. This method is based on a smartphone platform that is excited by a purple light source to capture spontaneous fluorescence images of skin and teeth. The Wiener estimation algorithm was used to reconstruct the acquired autofluorescence image into a multispectral data cube with 31 bands (400-700 nm). Furthermore, the spectral intensity of the porphyrin produced by the bacteria and the endogenous background tissue emission was extracted and compared. Based on the extracted autofluorescence spectra, a weighted subtraction method was utilized to obtain high-contrast and high-signal-to-noise identification of the bacteria on the skin surface. The experimental results indicate that the proposed method accurately reconstructs the multispectral fluorescence image and realizes the position of bacteria. After the weighted subtraction method was applied, the contrast increases by 12 to 46 times and the signal-to-noise ratio increases by 1.63 to 2 times.