The application of camera systems based on charge accumulating sensors for quantitative measurements of light distributions requires correction of significant systematic errors like non-linearity. A characteristic type of signal non-linearity that often occurs is a linear decreasing one. This contradicts the assumption of imperfections of the device to be the source of non-linearity but looks like a systematic effect. This work will describe the search for the reason of this type of non-linearity which ends in the observation of a spectral dependence of the non-linearity. A first approach to correct this effect without knowing the source spectrum is presented.