Accuracy issue from Firefly MV in Photosphere

All,

I have been creating HDR images for purposes of pixel-based luminance
analysis from photographs captured with a PointGrey Firefly MV [FMVU-13S2C
(Color)].

http://www.ptgrey.com/products/fireflymv/fireflymv_usb_firewire_cmos_camera.
asp

It is a board level camera with usb 2 connection, uses the Sony IMX035LQR-C
cmos sensor and a rolling shutter. I have fixed exposure to 1, brightness
to 0, gain to 0, and turned the frame rate off. I have then run multiple
exposures (10-12) using shutter speeds from 1ms to 8000ms excluding
completely saturated exposures. I have a fisheye lens attached but have
not been able to run vignetting correction yet since I do not trust my
response curve. I have been running accuracy comparisons to my Canon which I
trust with poor comparisons. It is successfully resolving in Photosphere,
but only after I manually write the EXIF header since the camera actually
does not write an EXIF header. I have written camera make, model, f stop,
and shutter into header.

Any ideas?

I would be happy to ftp exposures etc if it would help.

Thanks for any help!

Kevin Van Den Wymelenberg
Integrated Design Lab ­ Boise, Director
College of Art & Architecture, Assistant Professor
University of Idaho ­ Boise Center
www.uidaho.edu/idl

ph. 208.724.9456 fx. 208.343.0001 306
S. 6th Street Boise, ID 83702

Hi Kevin,

Are you errors in absolute terms or relative (e.g., one patch to another on a Macbeth chart or similar)?

-Greg

···

From: Kevin Van Den Wymelenberg <[email protected]>
Date: June 17, 2011 1:19:17 PM PDT

All,

I have been creating HDR images for purposes of pixel-based luminance analysis from photographs captured with a PointGrey Firefly MV [FMVU-13S2C (Color)].

http://www.ptgrey.com/products/fireflymv/fireflymv_usb_firewire_cmos_camera.asp

It is a board level camera with usb 2 connection, uses the Sony IMX035LQR-C cmos sensor and a rolling shutter. I have fixed exposure to 1, brightness to 0, gain to 0, and turned the frame rate off. I have then run multiple exposures (10-12) using shutter speeds from 1ms to 8000ms excluding completely saturated exposures. I have a fisheye lens attached but have not been able to run vignetting correction yet since I do not trust my response curve. I have been running accuracy comparisons to my Canon which I trust with poor comparisons. It is successfully resolving in Photosphere, but only after I manually write the EXIF header since the camera actually does not write an EXIF header. I have written camera make, model, f stop, and shutter into header.

Any ideas?

I would be happy to ftp exposures etc if it would help.

Thanks for any help!

Greg,

Thanks for the reply. My errors are in absolute terms. I have a Minolta
luminance meter and a Macbeth grey card in the middle of the scene. I am
also doing comparisons of absolute luminance values produced by the FireFly
as compared to the luminance results from my Canon camera on images captured
at the same time of the same scene. I have high confidence in my Canon
results from previous validations. The HDR that Photosphere resolves for
the Firefly at first look reasonable within the scene ­ it has logical
variation in luminance values. It is the absolute errors that I am worried
about.

Thanks!

Kevin Van Den Wymelenberg
Integrated Design Lab ­ Boise, Director
College of Art & Architecture, Assistant Professor
University of Idaho ­ Boise Center
www.uidaho.edu/idl

ph. 208.724.9456 fx. 208.343.0001 306
S. 6th Street Boise, ID 83702

···

From: "Gregory J. Ward" <[email protected]>
Reply-To: High Dynamic Range Imaging <[email protected]>
Date: Fri, 17 Jun 2011 13:44:13 -0700
To: High Dynamic Range Imaging <[email protected]>
Subject: Re: [HDRI] Accuracy issue from Firefly MV in Photosphere

Hi Kevin,

Are you errors in absolute terms or relative (e.g., one patch to another on
a Macbeth chart or similar)?

-Greg

From: Kevin Van Den Wymelenberg <[email protected]>

Date: June 17, 2011 1:19:17 PM PDT

All,

I have been creating HDR images for purposes of pixel-based luminance analysis
from photographs captured with a PointGrey Firefly MV [FMVU-13S2C (Color)].

http://www.ptgrey.com/products/fireflymv/fireflymv_usb_firewire_cmos_camera.as
p

It is a board level camera with usb 2 connection, uses the Sony IMX035LQR-C
cmos sensor and a rolling shutter. I have fixed exposure to 1, brightness to
0, gain to 0, and turned the frame rate off. I have then run multiple
exposures (10-12) using shutter speeds from 1ms to 8000ms excluding completely
saturated exposures. I have a fisheye lens attached but have not been able
to run vignetting correction yet since I do not trust my response curve. I
have been running accuracy comparisons to my Canon which I trust with poor
comparisons. It is successfully resolving in Photosphere, but only after I
manually write the EXIF header since the camera actually does not write an
EXIF header. I have written camera make, model, f stop, and shutter into
header.

Any ideas?

I would be happy to ftp exposures etc if it would help.

Thanks for any help!

_______________________________________________
HDRI mailing list
[email protected]
http://www.radiance-online.org/mailman/listinfo/hdri

If your relative proportions are accurate, that means Photosphere is doing a good job of solving for the system response. However, there's no way for it to solve for the absolute calibration. It just uses a value that is approximate based on ISO, which it sounds like you don't know. Select a region you've measured and use the "Apply" button to calibrate against your measured luminance. Use the dialog option to save this factor for your camera and you should be good from then on. (You'll need to be consistent about the camera make, model, and version in your added Exif header, though.)

-Greg

···

From: Kevin Van Den Wymelenberg <[email protected]>
Date: June 17, 2011 2:55:00 PM PDT

Greg,

Thanks for the reply. My errors are in absolute terms. I have a Minolta luminance meter and a Macbeth grey card in the middle of the scene. I am also doing comparisons of absolute luminance values produced by the FireFly as compared to the luminance results from my Canon camera on images captured at the same time of the same scene. I have high confidence in my Canon results from previous validations. The HDR that Photosphere resolves for the Firefly at first look reasonable within the scene – it has logical variation in luminance values. It is the absolute errors that I am worried about.

Thanks!

Greg,

Thanks - that makes sense. I just checked and the relative proportions of
the Firefly HDR are substantially different than the relative proportions of
the Canon HDR - bummer. Looks like I will need to figure out a way to
calculate ISO to input into the header. What does Photosphere need in the
EXIF header for best accuracy?

(Make and model)
Shutter speed
F stop
ISO

Anything else?

Kevin Van Den Wymelenberg
Integrated Design Lab ­ Boise, Director
College of Art & Architecture, Assistant Professor
University of Idaho ­ Boise Center
www.uidaho.edu/idl

ph. 208.724.9456 fx. 208.343.0001 306
S. 6th Street Boise, ID 83702

···

From: "Gregory J. Ward" <[email protected]>
Reply-To: High Dynamic Range Imaging <[email protected]>
Date: Fri, 17 Jun 2011 15:41:26 -0700
To: High Dynamic Range Imaging <[email protected]>
Subject: Re: [HDRI] Accuracy issue from Firefly MV in Photosphere

If your relative proportions are accurate, that means Photosphere is doing a
good job of solving for the system response. However, there's no way for it
to solve for the absolute calibration. It just uses a value that is
approximate based on ISO, which it sounds like you don't know. Select a
region you've measured and use the "Apply" button to calibrate against your
measured luminance. Use the dialog option to save this factor for your
camera and you should be good from then on. (You'll need to be consistent
about the camera make, model, and version in your added Exif header,
though.)

-Greg

From: Kevin Van Den Wymelenberg <[email protected]>
Date: June 17, 2011 2:55:00 PM PDT

Greg,

Thanks for the reply. My errors are in absolute terms. I have a Minolta

luminance meter and a Macbeth grey card in the middle of the scene. I am also
doing comparisons of absolute luminance values produced by the FireFly as
compared to the luminance results from my Canon camera on images captured at the
same time of the same scene. I have high confidence in my Canon results from
previous validations. The HDR that Photosphere resolves for the Firefly at
first look reasonable within the scene ­ it has logical variation in luminance
values. It is the absolute errors that I am worried about.

Thanks!

_______________________________________________
HDRI mailing list
[email protected]
http://www.radiance-online.org/mailman/listinfo/hdri

If your proportions are wrong in the result, adding information to the Exif header won't fix the problem. In general, Photosphere requires:

ASA
Speed
Aperture
Make
Model
Version

You need to make sure the white balance and sensitivity are fixed for all exposures, and that the camera is not messing around with the tone curve as many like to do these days.

-Greg

···

From: Kevin Van Den Wymelenberg <[email protected]>
Date: June 17, 2011 5:07:45 PM PDT

Greg,

Thanks - that makes sense. I just checked and the relative proportions of the Firefly HDR are substantially different than the relative proportions of the Canon HDR - bummer. Looks like I will need to figure out a way to calculate ISO to input into the header. What does Photosphere need in the EXIF header for best accuracy?

(Make and model)
Shutter speed
F stop
ISO

Anything else?