In this paper, we develop a methodology to extract the neural components from Brainbow images. Brainbow, a new genetic engineering technology, can help scientists understand the mechanism of an olfactory system. However, crosstalk among channels exists in these images. In addition, it is necessary to develop an automatic method to analyze these data. In our proposed system, we adopt Gaussian mixture model to model the phenomenon of crosstalk and reconstruct the Brainbow images. Spectral matting is applied to extract neural components from different channels. Under this system, we can extract useful information from Brainbow images efficiently and correctly.