Thread: technical image processing Q

So maggots know everything, right?

Say I have roughly 1000 images of part of a mountain. Some points in the images are of particular interest and want to find a way to get grayscale values for those points. The camera is mounted so that it may shake slightly in strong winds or for whatever reason, thus the images are not aligned perfectly.

I figure I need to get them aligned before I can do anything about the specific points. What would be a good way to approach this? Doing it manually is not an option, I need some sort of script. Any ideas?

I am used to working with Matlab and I think there are ways to use it for this, however I am only coming up with complicated ideas and I think there might be a simpler way to do this...

Maybe something like this using Photoshop's Merge to HDR functionality? You probably couldn't do 1000 at a time, but you could batch it.

http://www.blogarithms.com/index.php.../12/hdr-align/

Maybe PM StanWagon http://www.tetongravity.com/forums/m...7629-StanWagon

What exactly are you trying to extract? Do you really need pixel level detail? One possibility would be to use either erode or dilate (depending on if the area of interest is a low or high value). That would give you a bit more wiggle room at the expense of exact details. I'd work in Matlab if I were you, do you have the image processing toolbox?

Thanks for the input!

The photoshop approach is admittedly kind of tempting but I think it would be a huge pain in the long run (like after 20 pictures) and my supervisors would not approve because of lacking nerd factor.

geomorph: I do have the image processing toolbox and am looking for a version with the computer systems vision toolbox because that seems cool. There are a bunch of temperature loggers in a rocky slope and I am supposed to find a way to say if there is snow on the loggers in any given image or not, short lived summer snowfalls being most important (camera takes pictures every two hours). I was thinking of defining small areas around the loggers and averaging over them and setting a threshhold of some sort. the attached picture is a different place but that is the general sort of setting. Another issue is that I need to eliminate unusable images (fog, darkness etc)...

(ps: like my rock glacier?)

I think your approach sounds solid. I would define an ROI for each sensor, dilate the image to expand the areas of snow, and then average the grayscale values within the ROI and set a threshold for when there is snow (high values) or dark (low values). That looks like a nice field area. Is that in the Alps somewhere?

yes, Tyrol, Austria, Ötztal Alps..