부산대학교 도서관

Image sensors in cameras that are used for outdoor imaging, such as for automotive applications, are typically operated in the high dynamic range (HDR) mode. Photo-response linearity is a key performance indicator of image sensors. Existing procedures for the evaluation of image sensor linearity are based on exposure variation. They are efficient when image sensors are operated in the basic mode, which has a limited dynamic range (DR), however, they are difficult or impractical to implement with image sensors in the HDR mode. This work presents a method to evaluate the linearity of image sensors in the HDR mode. The method uses an indirect approach, which is based on the relative response between color channels in color cameras. It is independent of the HDR mechanism and the actual photo-response ratios between color channels; it can be applied to monochrome cameras with attenuation filters. We demonstrate the method with a camera module that includes a color image sensor that is activated in the HDR mode with the recommended configuration file and with a file that was modified to produce a nonlinear HDR response. Raw image data were collected with test targets in a controlled laboratory environment and outdoors in a representative HDR scene. The data were used to calculate signal and noise properties to construct the signal-to-noise ratio and relative channel response curves and to evaluate color properties. Analysis of the relative response curves shows that the proposed method is highly sensitive to deviation from linearity and that it correlates well with color artifacts.