Color Quantification in Endoscopic Video Sequences: Potentials and Limits

This paper compares color imaging principles and techniques of color quantification symbolizing the scheme of reality, human vision, and computer vision in analogy to the additive primaries red, green, and blue, respectively. The certain but unknown reality is not covered completely neither by human nor by computer vision. Up to now, most image processing techniques are designed to fit computer to human vision. Color transforms into perceptional color spaces such as Luv exemplify this common strategy. Derived from this model, those concepts are limited in general with respect to the grade of accordance of computer vision and reality, which is decisive for computer assisted diagnosis. Therefore, we suggest to move computer vision directly towards reality. The potentials of this strategy are demonstrated by color texture analysis as well as vibration measurements of vocal folds resulting in new quantitative parameters for video-laryngoscopic diagnostics. In other words, the future of color quantification in endoscopic video processing is seen in computing novel parameters rather than reproducing human vision.