The Influence of Bayer Demosaic Algorithms on Spatiotemporal Frequency Response Measurements

Abstract

The slanted edge method (e-SFR), as defined in ISO12233, is a popular method for quantifying camera spatial frequency response (SFR), because of its relative robustness against image noise and geometric distortion, efficient space utilization, and ability to measure beyond the Nyquist sampling limit of the sensor. Although e-SFR is known to be significantly influenced by image processing, few studies have explored how Bayer demosaic algorithms affect e-SFR measurements. In this study, we performed a through-focus sweep in 80 steps, capturing 30 raw Bayer pattern images at each step. We then compared SFR measurements from several of the most widely available Bayer pattern demosaic algorithms and compared the results with SFR measurements from raw Bayer pattern images. The results of the study show that of the demosaic algorithms tested, the Malvar et al. algorithm had the highest SFR@Ny/4, while the Menon et al. demosaic algorithm had the highest SFR at Ny/2 and at the Nyquist frequency. The results of the through-focus study showed that the choice of the demosaic algorithm can change the location of the measured peak focus step. This finding has significant implications for fixed-focus camera manufacturing, as a shift in the measured peak-focus position can result in a reduction of camera resolution. SFR measurements from white-balanced raw Bayer images had less measurement variance than those from demosaiced images. When characterizing camera focus, it is recommended to measure SFR using white-balanced raw Bayer pattern images for quality assurance, as it avoids image processing-related complications and has low measurement variability.

Keywords

• Color filter array
• demosaic
• e-SFR
• image quality
• ISO12233
• spatial frequency response