Worried with geometric distortions

[Mfranquelo]

Im worried with geometric distortions between two scans of the identical model, the differences are quite big... any possible reasons for this ?

Im using 35mm lenses.. and im not applying any geometric distortions corrections to my images nor cropping...

Thank you!

[Mr_Curious]

Im worried with geometric distortions between two scans of the identical model, the differences are quite big... any possible reasons for this ?

Im using 35mm lenses.. and im not applying any geometric distortions corrections to my images nor cropping...

Thank you!

Wow... that's quite a big difference!

[Wishgranter]

hmm just curious, have not just enabled ORTHO/Perspective ?? can show Aligment of the  dataset ??

[Mfranquelo]

That was what i thought wishgranter, but i've double checked many times and both are on perspective view.

Im re-aligning the one that looks wider... I'll post more images tomorrow, but this is driving me crazy!

[Mfranquelo]

By the way i've just came up with something i may did wrong... I resized the bounding box "resize region" after the alignment process.... this may be the cause ? 

[Wishgranter]

NO, BB is just region of reconstruction... can send me both datasets ? And the  distrtion is clearly tosee in AGI orjust  after postprocess in Zbrush ??

[Alexey Pasumansky]

For us it will be also interesting to take a look at the project files.

Maybe you can also post screenshots from PhotoScan model view with the models of the same scale (for example, by aligning the chunks based on cameras)?

[Mfranquelo]

They're distorted on Agisoft as well im afraid..., i?ll upload the project files to Wetransfer.
Thank you for your help 
 

[Mfranquelo]

Please download the project through this link: http://we.tl/HI5omMWNZG

Chunk1: First scan 
Chunk3: Second scan

Both chunks have been aligned by points: High accuracy.

It comes with source exported JPGE files as well. Raw's have been processed in Lightroom just bumping up the shadows slightly.
Camera's : Canon 700d
Lenses: Samyang 35mm 50mm equiv    ** This lens doesnt provide information to the EXIF, however if im correct, by default Agisoft assumes 35mm which is fine.
Capturing time: 3 seconds. (Using my 8 camera rotating semi circular frame)
48 images.


** I can try to scan known size objects, like squares or an sphere several times, to test inconsistencies if its necessary.
Let me know what you think.

[Wishgranter]

Hi, yes just prepare some well know and measured object, that can be beasured, so the problem can be pinpointed.....

[Wishgranter]

Hmm seems that agisoft cannot handle correctly the  LENS - focal lenght.....

[David Cockey]

The actual difference in the PhotoScan models is small.

I downloaded the files and opened the model. Created low density point clouds for each chunk. Aligned the chunks based on points (low accuracy). Looked at the aligned point clouds together. Not much difference. Created meshes and decimated each to 10,000 faces. Looked at the mesh structures together. Not much difference.

Differences in the subject between the two sets of photos include:
- Jaw and lip positions
- Amount eyes are open
- Position of neckline of clothing
- Hair and position of headband
- Amount of makeup
- Skin blemishes

[Wishgranter]

can createa screenshots of the models  ( Mesh ), it could be problem of Zbrush, get sma eproblem few months back..... try export camera Calibration and compare results, but model seem to be the same....

see fast compare in pshop https://dl.dropboxusercontent.com/u/15047343/Facial.jpg

[David Cockey]

Im worried with geometric distortions between two scans of the identical model, the differences are quite big... any possible reasons for this ?

Im using 35mm lenses.. and im not applying any geometric distortions corrections to my images nor cropping...

Thank you!

The models are not "identical". See the list in my post above of the differences. Any comparison requires alignment. The differences in the photos in the original post are due in part to different camera positions.

Edit: Another possible difference is the position of the head relative to the torso.

[Mfranquelo]

I will check this again with the models aligned on Agisoft and mesh built. Im gratefull for your input guys. 

- You're totally true David, i know that they are not identical, i've scanned the first one before a botox inyection and the other one after (presenting several traumas) as well as the expression modified of course. But my worrying came when i saw that the proportions between both of them, werent more or less identical.

"The differences in the photos in the original post are due in part to different camera positions."

But both of the models are practically with the same camera perspective arent they ? If you have a look the bottom of the ears align perfectly with the mouth in both of them. How can a little camera missalignment produce so different results ?

* I'll check now the models on Agisoft after alignment and get back !

Thank you.