Forum

Author Topic: Deformed scan  (Read 2657 times)

deadwilder

  • Newbie
  • *
  • Posts: 5
    • View Profile
Deformed scan
« on: April 25, 2018, 07:13:23 AM »
Hello, I've had a couple of recent scans come out a bit deformed, despite mostly looking like the scanned objects. I've noticed the height of the scan seems correct, but the width is smaller than of the original object, making the scan look deformed. I'm trying to understand if I'm photographing the object incorrectly or if there is any setting in PS that I'm not properly setting up.
I'm shooting RAW on a turntable and light tent with a Canon 5D Mk II, using a 100mm macro lens, 32 f/stop, 1.3 sec, on a tripod, with a flash ring and two flash lights. I haven't edited the RAW files before processing them on PS, I exported them as high-quality JPGs.
I've attached here an image of the scan as well as a couple of samples of the photographs.
Any tips would be greatly appreciated.
« Last Edit: June 22, 2018, 06:41:48 PM by deadwilder »

deadwilder

  • Newbie
  • *
  • Posts: 5
    • View Profile
Re: Deformed scan
« Reply #1 on: April 25, 2018, 07:26:37 AM »
More samples of the photographs I took attached here.
« Last Edit: June 22, 2018, 06:42:10 PM by deadwilder »

Jobbo90

  • Newbie
  • *
  • Posts: 19
    • View Profile
Re: Deformed scan
« Reply #2 on: April 25, 2018, 02:12:44 PM »
Are you using GCP points? If so, might be worth to check if the coordinates you use are correct, else it forces the model in an incorrect projection causing deformations in the final object..

deadwilder

  • Newbie
  • *
  • Posts: 5
    • View Profile
Re: Deformed scan
« Reply #3 on: April 25, 2018, 05:49:10 PM »
Are you using GCP points? If so, might be worth to check if the coordinates you use are correct, else it forces the model in an incorrect projection causing deformations in the final object..

I’m not using markers or GCP points. My workflow is to import the images; disable the ones that are not great; apply mask if necessary; align photos; build dense cloud, mesh and texture.

In the case of this model it wasn’t necessary to apply a mask since the only background information in the photos was mostly the scale beneath the object.
« Last Edit: April 25, 2018, 05:50:53 PM by deadwilder »

Jobbo90

  • Newbie
  • *
  • Posts: 19
    • View Profile
Re: Deformed scan
« Reply #4 on: April 26, 2018, 01:44:29 PM »
Are you using GCP points? If so, might be worth to check if the coordinates you use are correct, else it forces the model in an incorrect projection causing deformations in the final object..

I’m not using markers or GCP points. My workflow is to import the images; disable the ones that are not great; apply mask if necessary; align photos; build dense cloud, mesh and texture.

In the case of this model it wasn’t necessary to apply a mask since the only background information in the photos was mostly the scale beneath the object.

Then it might be worth to implement a arbitrarty reference system with some control points? Use the squares of the underlying surface with elevation of 0 and a x and y coordinate from an arbitrary chosen 0,0? See if it helps to force the model in correct dimensions. Else you might have to describe in some more details your workflow in order to be able to answer this question..


James

  • Hero Member
  • *****
  • Posts: 756
    • View Profile
Re: Deformed scan
« Reply #5 on: April 27, 2018, 10:36:22 AM »
Then it might be worth to implement a arbitrarty reference system with some control points? Use the squares of the underlying surface with elevation of 0 and a x and y coordinate from an arbitrary chosen 0,0? See if it helps to force the model in correct dimensions. Else you might have to describe in some more details your workflow in order to be able to answer this question..

I thought of that too, but it looks from the 'deformed' screenshot that the grid at the base is proportionally correct, and that it is the 'Z' that is relatively stretched up.

The image showing some of the sample images shows that they are mostly looking downwards, and with a long (ish) lens like a 100mm you will get greater inaccuracy along the camera axis due to reconstruction uncertainty, which may be why Z is distorted more than XY.

One way to test/fix that would be to include some photos looking horizontally at the subject, or add some features with known height above the grid, like small wooden cubes with the same grid pattern on to enable GCPs to be added with variation/distribution in all 3 dimensions.

Alternatively, just a couple of objects with known height so that scale bars could be added in both horizontal and vertical axes.