Estimating spectral reflectance in spectral imaging system

摘要

The identification of tumor is important for the diagnosis of early cancer. Optical reflectance spectrum of human tissue could reveal the characteristics of the tissue that helps medical diagnosis. Multispectral imaging techniques have been widely used for the application of digital recording. In this article we developed a multispectral imaging system with a monochrome CCD, five optical filters including conventional RGB filters and two narrow band filters, light source and imaging processor. We propose a method to reconstruct the reflectance spectrum using a selective combination of spectral estimation technique. With the reconstructed spectrum, we can form the image of the object in certain wavelength that is not in the range of any of the wavelengths corresponding to the above filters. The imaging constructing speed of multispectral imaging system is limited by CCD speed. In order to reconstruct the reflectance spectrum in real-time, the number of the filters involved in the reconstruction should be as less as possible. In the experiment, we investigate the relationship between the number of the filter channels and the estimation error (RMSE).. To minimize the estimation error, we use Weiner estimation to choose the best combination of filter channels for certain wavelength. Using the proposed method, we can achieve high accuracy of image reconstruction with limited number of filter channels in real time.In this paper, the multispectral imaging system setup is briefly introduced, and the method of reconstructing the reflectance spectrum is described.