| Abstract |
Tricolor imaging systems are widely deployed as they produce satisfying results for
everyday use. However, for the use in industry, computer graphics, and design, it
is important to achieve an even higher level of precision. This can be reached via
multispectral images. Instead of three color values, each pixel of such an image contains a whole spectral power distribution.
The aim of this work is to motivate, describe, and build a multispectral imaging system using devices such as a high sensitive camera and a tunable fi lter. Similar systems have already been reviewed in the literature. The difficulties are to adapt such a system, since the complexity of the di fferent devices and models is high. Furthermore they have to be selected accordingly to the purpose of the system, which is in this case recording spectral images of natural scenes.
Di fferent devices were integrated in a common framework. Especially interesting in this context was the camera API and the process of examining and accessing the recording sequence and exposure settings which will be documented in this thesis. Its functionality had to be synchronized with the fi lter, in order to automatically record exposure sequences over di fferent wavelengths.
Based on the implementation, the system was calibrated through the use of a monochromator and an integrated sphere. In a series of experiments the light of the monochromator was recorded using a calibrated spectroradiometer and the multispectral imaging system. Through the recorded data the multispectral imaging system could be corrected for noise and hence its characterization, the spectral sensitivity, position of the measurements points on the wavelength axis, and normalization matrix, was achieved. The models were validated in a final experiment.
The result was the attainment of a fully integrated multispectral imaging system. For
this system the noise model and the spectral sensitivity are known factors. The finding about this synergy could be used for further research projects. |