I am trying to develop a simple method for amateurs to make relative measurements of noise in digital cameras. One method involves taking multiple images of a standard grey card, bringing the images into Photoshop as a stack and then using Photoshop's standard deviation function. As expected, the mean value of the standard deviation over the image increases with ISO setting. However, there is also a distinct spatial pattern to the standard deviation visible at all ISO settings. (See example image attached.) Changing lenses or just adjusting focus changes the pattern. No corresponding pattern is observed when calcuting mean or median.
My experiments were conducted on a copy stand using a Canon 5D Mark II DSLR and a Kodak grey card. The attached image is the pattern when the card is in focus. The contrast has been enhanced to make the pattern clearer. The standard deviation was calculate from 10 images taken in rapid sequence under close to identical conditions.
The ring pattern may be Fresnel zones for the optics. I think they may appear for analogue cameras too. Although a periodic sensor may open up for moire effects if there are other periodic phenomena at action.
Anders Buen Interesting. Any idea why this would show up in the variance and no in the mean?
Douglas Stinson I do not know. A file in photoshop, and the screen is periodic. Taking a picture of a screen may give moire patterns. I guess that would be the mean, just one pic. I havent experimented with picture averaging and optics in practice. But use statistics a lot on data in general.
I do have an optics course from the study time and worked with a large laser for TV holography recordings. The small lenses we used would always give fringe patterns doe to imperfections in the glass or small particles. The lenses were used to spread the laser beam over an object for study. It would not look as periodic as your pic, but still there would be irregularl cocentric fringes with shifting maxima and minima.