Sharpness and focal length

[gregoconn]

I've been trying to research this as much as I can before posting this but I haven't found as much information as I would have liked specifically regarding photoscan.

Currently I have invested in a Canon 60D and a 24-105mm lense for recreational purposed. I'm now using this camera for Photoscan and would like to invest in a prime lens. From what I understand a 50mm lense on a full frame sensor is the best way to go to get great results.

I'd like to know what the best set up for a crop sensor is. I know that the sigma f1.4 30mm equats to ~48mm and the Canon f2.0 35mm  works out to be ~55mm. Mean while your 50mm lenses come to about 70mm. Which is more appropriate? Does it matter? Personally I would like to go the Sigma 30mm or Canon 35mm as I have limited room, but if it's going to hurt  my results I'd rather spend the money on a 50mm.

Also a question on sharpness. Would the difference in sharpness between a Canon f1.8 50mm @ f8 and a Canon f1.4 50mm @ f8 make a noticable difference to my results?

http://www.astrosurf.com/buil/50mm/test_us.htm

I'm hoping to get a bit of clarification on this before I start spending money on the lenses and testing with my 24-105 is difficult because it doesn't have a zoom lock and the focal length slips over time (v frustrating)

Also any recommendation for lenses would be much appreciated.

Cheers,

Greg

[jedfrechette]

testing with my 24-105 is difficult because it doesn't have a zoom lock and the focal length slips over time (v frustrating)

A little strip of Pro-gaff is often a pretty effective solution to this particular problem.

[gregoconn]

Hmmm..good suggestion. Not sure how keen I am to gaffer my lens though. I hear you can tighten some screws under the zoom ring as well which helps but again, not too keen on messing with my lens for now.

I'd still be keen to hear other peoples thoughts about focal lengths with a crop sensor.

[tincansassoc]

Tape is a good solution.

As far as lenses go I have been trying a Canon 16-35 L lens on a cropped sensor 450d and am getting decent results, but I am also uncertain about how wide angle of a lens is too wide for photoscan. I have been shooting the photos in about the 18 - 25mm range and have been trying to scan various non-planar objects that are a meter or less in size. I am having issues on some models with holes in my geometry, however I believe this is more due to how the photos were taken than about how wide the lens is. I have had some good results with some simple cube-shapped objects which were almost all shot near the 16mm range with the cropped sensor camera. That particular model only used about 40 photos, but a similar model using about 70 photos had some issues with holes in the model. I am finding that fine tuning the settings for each model makes a big difference...

Since your lense is similar build quality to mine and we have overlapping focal lengths, I believe it would be safe to say your lens is fine. You may even have an advantage by having the IS lens because photoscan appears to work much better when you use high f-stops and low ISO, which requires you to have a lot of light to not get blury photos if your shooting handheld.

Personally I would like to know if people here have better results with an IS lense (or VR for Nikon users) since there are now some prime lenses with IS. Has anyone tried a lense such as the Canon 28mm IS with photoscan?

[jedfrechette]

In general, you want to stay away from image stabilization for photogrammetric applications. If IS is turned on your camera calibration parameters will be slightly different for each image. Essentially, you're stabilizing the image by moving the principal point around.

One other note on tape, as long as you are using a good quality tape that isn't going to leave a bunch of residue behind on your lens I don't think there is much to worry about.

[gregoconn]

Excellent. That's good to know. I've been shooting my photo's currently with IS on. I'll give the tape a go and do a few tests at 30, 35 and 50.

I'm still curious about if the extra sharpness from the more expensive lenses would help. I'm always tempted to go for the expensive glass, but when you start talking about buying multiple camera's. 1 L glass = 3 cheapies.

Has anyone tried multiple camera's using different lenses? Would it hurt your results using a 1.2 50mm @f8 on what camera and 1.8 50mm @f8 on another? I've tried 1 handheld with multiple focal lengths and that really doesn't solve well.

[tincansassoc]

Thanks for the IS comment, this makes good sense.

It has been my understanding that as long as your images are un-edited and still retain all their exif camera data, then mixing cameras, lenses or focal lengths should not be a problem. Assuming you stick within a certain range of focal lengths. I have read some posts on this forum that suggest mixing cameras is not a problem. Although I would also like to hear what focal lengths people have been most successful using and with what camera body.

[RalfH]

I don't see a problem in using zoom and/or IS lenses or even different cameras - PhotoScan models lens parameters very well except for extreme wide angle lenses. The important thing when using lenses in which optical components are moving between shots (zoom, image stabilisation, focussing) is that camera calibration settings should be set to "separate" instead of "groups" as this allows PhotoScan to use a different calibration for each image. The exception from this is when you want to use a high-end, very stable lens (fix focal lens, no focussing between images etc.) where you might want to create a very good calibration and use that same calibration for all images. But generally, you don't need to worry about these things.