Measuring Noise and the Influence of Noise Reduction on Resolution
Digital imaging system resolution is determined by a combination of sensor characteristics, lens characteristics, and image-processing algorithms. As pixel size decreases, sensitivity decreases and noise increases, requiring a more sophisticated noise-reduction algorithm to obtain good image quality. Furthermore, small pixels require high-resolution optics with low chromatic aberration and very small blur circles. Ultimately, there is a tradeoff between noise, resolution, sharpness, and the quality of an image.
Although noise is typically measured as a standard deviation in a patch with uniform color, it does not always accurately represent human perception. Based on the "visual noise" algorithm described in ISO 15739, an improved approach for measuring noise as an image quality aspect will be demonstrated. The course shows a way to optimize image quality by balancing the tradeoff between noise and resolution using sample images.
Speaker: Dietmar Wueller
Dietmar Wueller studied photographic sciences at the University of Cologne. He owns a test lab for digital photography and has been testing digital cameras and scanners for German magazines and manufacturers since 1997. He is the editor of the ISO scanner standards (ISO 21550 and ISO 16067) and the vice chairman of the photography section in the German DIN. He also chairs the digital photography working group in the European Color Initiative (ECI).
Google Tech Talks
January 22, 2009