På svenska

Presentation Information     2009-02-16 (15:15)   •  The seminar room at Vi2

Speaker Hamid Sarve
Title Color Image Processing in FPGA
Abstract This thesis describes the development of a low-error real-time FPGA realization of the reconstruction and subsequent transformation of Bayer-filtered images into the CIE-L*a*b* color space.

Algorithms for recovering the lost color information of Bayer filtered images, so called demosaicing algorithms, were evaluated in their ability to yield a minimal reconstructional error. It was shown that the Adaptive Color Plane Interpolation demosaicing algorithm generates the lowest error at the expense of hardware resources.

The images were further more converted into the perceptually balanced and device independent color space CIE-L*a*b*. Methods for approximating the cubic root function of the transformation were discussed. A relatively precise approximation was obtained using two Look-Up Tables.