gmane.technology.imaging.hdri

Re: hdrgen and photosphere

Gregory J. Ward — 2013-05-14T14:41:48

Erik's post came in after Mehlika's because he got caught in the moderator trap -- his e-mail was not "on the list."

My apologies for not answering the original post, but I was on a plane or two....

Cheers,
-Greg

Re: hdrgen and photosphere

Erik Reinhard — 2013-05-13T21:24:48

Hi All,

I am not entirely sure that I remember this correctly, but I once looked at these
camera curves in photosphere for the purpose of plotting them. I seem to recall
that the coefficients are reversed relative to how you might expect them.

It may be worth trying to reverse the order of the coefficients, so for the first of
the three lines this would be:

4 4.462320e+00 -5.786777e+00 2.441168e+00 -3.556814e-02 7.135383e-03

Hope this helps.

Cheers,
Erik



On May 13, 2013, at 10:31 PM, Axel Jacobs <jacobs.axel-Re5JQEeQqe8AvxtiuMwx3w< at >public.gmane.org> wrote:

Re: hdrgen and photosphere

Michael Martinez — 2013-05-14T00:32:13

Ah ha!! That's the trick... Things look great now. 

Many thanks, both to Axel and Mehlika. 

MM






On May 13, 2013, at 3:36 PM, Mehlika Inanici wrote:

Re: hdrgen and photosphere

Mehlika Inanici — 2013-05-13T22:36:25


As I recall, the order is reverse between Photosphere and hdrgen. If the R response is from your Photosphere file is
7.135383e-03,-3.556814e-02,2.441168e+00,-5.786777e+00,4.462320e+00

it should be a "mirror image" in hdrgen:
4.462320e+00, -5.786777e+00, 2.441168e+00, -3.556814e-02, 7.135383e-03

Mehlika


On Mon, 13 May 2013, Michael Martinez wrote:

Re: hdrgen and photosphere

Michael Martinez — 2013-05-13T22:23:53

The rsp looks good - nice and smooth, which makes sense as Photosphere seems to have no trouble with these values. I also tried straight decimal format, and still no luck...

Very curious...  




On May 13, 2013, at 1:31 PM, Axel Jacobs wrote:

Re: hdrgen and photosphere

Axel Jacobs — 2013-05-13T20:37:23

PS: Two more thoughts:

I think hdrgen copes well with scientific notation, but just to 
double-check: try converting from scientific notation to straight decimal.

Also, I did notice with some of my won sequences that wood with its 
grain and large areas of non-grey can be tricky. Just for the 
calibration, I suggest you choose a grey background.

Axel



On 05/13/2013 09:21 PM, Michael Martinez wrote:

Re: hdrgen and photosphere

Axel Jacobs — 2013-05-13T20:31:28

Right, so we are on the right track--hdrgen no longer rejects the RSP.

Next question is: how good is the RSP? Try plotting it. If it is smooth 
and monotonic, without any kinks in it, it's a good one. Otherwise, try 
generating it from a different sequence. The 5D is used a lot on WebHDR:
http://www.jaloxa.eu/webhdr/cameras/Canon__Canon_EOS_5D.shtml
How does your RSP compare with the averaged one? Note that there is no 
quality control, and even 'wonky' ones make it into the average.

Axel



On 05/13/2013 09:21 PM, Michael Martinez wrote:

Re: hdrgen and photosphere

Michael Martinez — 2013-05-13T20:21:21

Using a 4 in front of the Photosphere numbers like this:

4 7.135383e-03 -3.556814e-02 2.441168e+00 -5.786777e+00 4.462320e+00
4 4.543790e-03  3.363440e-02 1.917460e+00 -4.660105e+00 3.792747e+00
4 3.956902e-03  3.244581e-02 1.827308e+00 -4.505138e+00 3.729707e+00

unfortunately results in a wonky HDR with garbage luminance values and strage visual artifacts - see screen shot here. 

compared to a screenshot of the photosphere created HDR





On May 13, 2013, at 1:05 PM, Axel Jacobs wrote:


_______________________________________________
HDRI mailing list
HDRI-mVArnbgP/ra1OXqM+dc5E0B+6BGkLq7r< at >public.gmane.org
http://www.radiance-online.org/mailman/listinfo/hdri

Re: hdrgen and photosphere

Michael Martinez — 2013-05-13T20:06:44

Hmm - pre-appending with a 6 didn't do the trick. 

It seems like hdrgen really wants a 3rd order polynomial (no matter what I put into the .rsp, hdrgen overwrites it as such), and I'm not too sure how to translate the 5 values for the R, G, and B channels generated by Photosphere. 

As another data point, the other camera curves in my Photosphere preferences file have a variety of values, either 4, 5 or 6 for each channel, depending on the camera. 

As a work around, I can apply a multiplier to the .rsp that hdrgen makes, and that gets me pretty close. But I'd also love to know if there's a more direct way to translate the Photosphere data into hdrgen friendly data. 

Thanks all - 

Mike


 

On May 13, 2013, at 12:14 PM, Axel Jacobs wrote:

Re: hdrgen and photosphere

Axel Jacobs — 2013-05-13T20:05:54

Oops,

try '4 ', not '6 ' as per my last post.

Embarrassingly yours

Axel

Re: hdrgen and photosphere

Axel Jacobs — 2013-05-13T19:14:55

Hi Michael,
 > I'm trying to use hdrgen with a response curve created by photosphere.

The first number in each line in an RSP file is the order of the 
polynome. Since it is missing in the RSP above (the one you hand-crafted 
from the Photosphere response), hdrgen deems it invalid and overwrites 
it with a new one. Try pre-pending a '6 ' to all three lines. hdrgen 
should than be happy and use the RSP that you supplied, rather than 
generate one from scratch.


More info here:
http://www.jaloxa.eu/webhdr/calibrate.shtml

Cheers

Axel

hdrgen and photosphere

Michael Martinez — 2013-05-13T18:44:19

Hi Folks -  

I'm trying to use hdrgen with a response curve created by photosphere. For the image, I have calibration points measured with a luminance meter, and when I make the hdr with photosphere, I get an accurate luminance measurement, but hdrgen gives me a value that's consistently ~10% higher. This is surprising as (I think) I'm using the same response curve for each process. 

Here's my hdrgen command: 
hdrgen -o ../01.hdr -r canon_5D.rsp -q 100 -f -g *.JPG

Here's what I put into canon_5D.rsp:
7.135383e-03 -3.556814e-02 2.441168e+00 -5.786777e+00 4.462320e+00
4.543790e-03  3.363440e-02 1.917460e+00 -4.660105e+00 3.792747e+00
3.956902e-03  3.244581e-02 1.827308e+00 -4.505138e+00 3.729707e+00

Here's canon_5D.rsp after hdrgen is run:
3 1.49643 -0.979 0.486712 -0.00413976
3 1.44009 -0.884479 0.448854 -0.00446749
3 1.46366 -0.875187 0.414554 -0.00303108

And here's the response curve from Photosphere:
"Canon"|"Canon EOS 5D Mark II"|"v.0"|{7.135383e-03,-3.556814e-02,2.441168e+00,-5.786777e+00,4.462320e+00}|{4.543790e-03,3.363440e-02,1.917460e+00,-4.660105e+00,3.792747e+00}|{3.956902e-03,3.244581e-02,1.827308e+00,-4.505138e+00,3.729707e+00}

Am I missing any hdrgen options, or perhaps making an error in my .rsp file? Thanks in advance for any insights. 

Mike



Michael Martinez ~ Associate
LOISOS + UBBELOHDE
- - - - - - - - - - - - - - - - - - - - - - - - - - - 
1917 Clement Avenue  Building 10A
Alameda, CA  94501 USA
- - - - - - - - - - - - - - - - - - - - - - - - - - - 
510 521 3800 VOICE
510 521 3820 FAX
- - - - - - - - - - - - - - - - - - - - - - - - - - - 
www.coolshadow.com





_______________________________________________
HDRI mailing list
HDRI-mVArnbgP/ra1OXqM+dc5E0B+6BGkLq7r< at >public.gmane.org
http://www.radiance-online.org/mailman/listinfo/hdri

Call for papers, Special session of PCS

Ward, Gregory J. — 2013-04-29T21:21:51

Please share with potentially interested colleagues....

Special HDR session at PCS 2013
The 30th Picture Coding Symposium (PCS 2013) Dec 8-11, 2013, San Jose, California, (www.pcs2013.org<http://www.pcs2013.org/> ) is the leading forum devoted specifically to advancements in visual data coding. This year there will be a special session on HDR. In particular we are looking for papers on:
- HDR still image encoding
- HDR moving image encoding

but will also consider topics such as:
- HDR display
- HDR still and video capture
- History and background of HDR
- HDR as it relates to cinema special effects and post-production

If you would like to be considered for this Special Session, please email a provisional title and short abstract by Wednesday 15 May.
The 4 page (max) submission is then due by 1 June. These submissions will be reviewed and between 4-6 selected to form the Session.

Best wishes
Greg Ward, Timo Kunkel and Alan Chalmers
greg.ward-BCqGbv6DZ10AvxtiuMwx3w< at >public.gmane.org<mailto:greg.ward-BCqGbv6DZ10AvxtiuMwx3w< at >public.gmane.org>
timo.kunkel-BCqGbv6DZ10AvxtiuMwx3w< at >public.gmane.org<mailto:timo.kunkel-BCqGbv6DZ10AvxtiuMwx3w< at >public.gmane.org>
alan.chalmers-OmadNMBGD7E2EctHIo1CcQ< at >public.gmane.org<mailto:alan.chalmers-OmadNMBGD7E2EctHIo1CcQ< at >public.gmane.org>

_______________________________________________
HDRI mailing list
HDRI-mVArnbgP/ra1OXqM+dc5E0B+6BGkLq7r< at >public.gmane.org
http://www.radiance-online.org/mailman/listinfo/hdri

Re: unknown view type

Chris Kallie — 2013-04-24T17:33:21

Greg confirmed my foggy memory about Photoshop and hdr headers. Just a 
couple quick thoughts I wanted to share... If you have extra time and 
energy, consider trying Greg's raw2hdr script 
(http://www.radiance-online.org/pipermail/hdri/2012-February/000363.html). 
However, there's really no problem with using jpegs and 
hdrgen/photosphere. I have personally found that with creating and using 
a good calibration method and file (see this: 
http://www.jaloxa.eu/webhdr/calibrate.shtml), and being consistent with 
the whitebalance setting, the jpeg method produces extremely accurate 
luminance images (r^2 > .996). As Greg just mentioned, a photometer to 
dial in the range would further improve the 1:1 hdr to real photometric 
values. Finally, I will cautiously note that in one experiment, I found 
that jpegs were more accurate than nefs (perhaps due to random 
fluctuations in estimating the calibration response curve), but I 
haven't repeated that experiment. For both image types (jpg and nef), 
the covariances between image sample and photometer were still extreme 
(r^2 ~ .99).

Happy imaging!
Chris

On 4/24/13 12:09 PM, Gregory J. Ward wrote:

_______________________________________________
HDRI mailing list
HDRI-mVArnbgP/ra1OXqM+dc5E0B+6BGkLq7r< at >public.gmane.org
http://www.radiance-online.org/mailman/listinfo/hdri

Re: unknown view type

Gregory J. Ward — 2013-04-24T17:09:49

Yes, Photoshop strips the header from Radiance (HDR) images, including the view parameters.  You can always add them back in afterwards using the Radiance "vinfo" command (on Unix anyway).  You'll still have the problem that Photoshop doesn't do any kind of absolute calibration on HDR images, so if evalglare needs absolute quantities, it won't have them.  Adding the correct exposure requires having an in-scene luminance measurement for calibration.

This is why it's generally better to use hdrgen or Photosphere for HDR creation in photometric applications, although a calibration measurement will improve your accuracy even more.

Cheers,
-Greg

_______________________________________________
HDRI mailing list
HDRI-mVArnbgP/ra1OXqM+dc5E0B+6BGkLq7r< at >public.gmane.org
http://www.radiance-online.org/mailman/listinfo/hdri

Re: unknown view type

Avinash Gautam — 2013-04-24T16:37:03

Hi Chris,

I think Adobe Photoshop must have messed up with the header files while creating the HDR. I was about to try the methods you sent in your earlier email but seems there is a shorter route by using hdrgen or photosphere to create the HDR file. I am a novice in this area and your help is really appreciated.

Thanks again for all your help.

Regards,
Avinash



From: Chris Kallie [mailto:kallie-OJFnDUYgAso< at >public.gmane.org]
Sent: Wednesday, April 24, 2013 10:28 AM
To: hdri-mVArnbgP/ra1OXqM+dc5E0B+6BGkLq7r< at >public.gmane.org
Subject: Re: [HDRI] unknown view type

Hi Avinash,

Just a followup: I was able to install and run evalglare on the .hdr created from last night's .nef files (Method 1). I then performed the following two operations:

pfilt -1 -x /4 -y /4 test01.hdr>test01s.hdr
evalglare -c test01g.hdr test01s.hdr

Everything seemed to work (i.e., I got a grayscale output image with a green 'glare' mask). I suspect that since the header information (I was calling it 'metadata' before) was already in the image, there was no need to add -vtv, or declare -vv -vh. Here is what I found in my test01s.hdr header:

#?RADIANCE
CAMERA= NIKON D90 version dcraw v9.12
hdrgen created HDR image from '_DSC3776.tif' '_DSC3775.tif' '_DSC3774.tif' '_DSC3773.tif' '_DSC3772.tif'
EXPOSURE=3.280407e+00
VIEW= -vtv -vh 37.200130 -vv 25.245402
CAPDATE= 2013:04:23 23:46:20
PRIMARIES= 0.6400 0.3300 0.3000 0.6000 0.1500 0.0600 0.3127 0.3290
FORMAT=32-bit_rle_rgbe
pfilt -1 -x /4 -y /4

I hope this helps...
-Chris

On 4/23/13 10:31 PM, Avinash Gautam wrote:
I am trying to generate a hdri image from my camera Nikon D90 using various lenses. My Nikon camera generates a .NEF file in RAW version however I was able to generate a .HDR file using .JPG of the same images in Photoshop. After I generated the .HDR file I tried running evalglare on it which gives me an error. I understand I need to have a fish eye lens to get the full view in order to get the correct results but the purpose of my experiment is to test the error in dgp using different lenses that are not fish eye lens. I am using the -vth command and have entered the -vv and -vh as calculated for the 35mm lens on a DX format camera (25.74degrees and 37.4degrees). This gives me an error pict_update_view: unknown view type and error: invalid view specified. Can anyone please let me know where I am going wrong.

Thanks,
Avinash







_______________________________________________

HDRI mailing list

HDRI-mVArnbgP/ra1OXqM+dc5E0B+6BGkLq7r< at >public.gmane.org<mailto:HDRI-mVArnbgP/ra1OXqM+dc5E0B+6BGkLq7r< at >public.gmane.org>

http://www.radiance-online.org/mailman/listinfo/hdri

_______________________________________________
HDRI mailing list
HDRI-mVArnbgP/ra1OXqM+dc5E0B+6BGkLq7r< at >public.gmane.org
http://www.radiance-online.org/mailman/listinfo/hdri

Re: unknown view type

Chris Kallie — 2013-04-24T16:28:25

Hi Avinash,

Just a followup: I was able to install and run evalglare on the .hdr 
created from last night's .nef files (Method 1). I then performed the 
following two operations:

pfilt -1 -x /4 -y /4 test01.hdr>test01s.hdr
evalglare -c test01g.hdr test01s.hdr

Everything seemed to work (i.e., I got a grayscale output image with a 
green 'glare' mask). I suspect that since the header information (I was 
calling it 'metadata' before) was already in the image, there was no 
need to add -vtv, or declare -vv -vh. Here is what I found in my 
test01s.hdr header:

#?RADIANCE
CAMERA= NIKON D90 version dcraw v9.12
hdrgen created HDR image from '_DSC3776.tif' '_DSC3775.tif' 
'_DSC3774.tif' '_DSC3773.tif' '_DSC3772.tif'
EXPOSURE=3.280407e+00
VIEW= -vtv -vh 37.200130 -vv 25.245402
CAPDATE= 2013:04:23 23:46:20
PRIMARIES= 0.6400 0.3300 0.3000 0.6000 0.1500 0.0600 0.3127 0.3290
FORMAT=32-bit_rle_rgbe
pfilt -1 -x /4 -y /4

I hope this helps...
-Chris


On 4/23/13 10:31 PM, Avinash Gautam wrote:

_______________________________________________
HDRI mailing list
HDRI-mVArnbgP/ra1OXqM+dc5E0B+6BGkLq7r< at >public.gmane.org
http://www.radiance-online.org/mailman/listinfo/hdri

Re: unknown view type

Chris Kallie — 2013-04-24T06:21:33

Hello Avinash,

There are two possible issues that I can think of: one regarding image 
size (more likely culprit), and the second regarding metadata (less 
likely). I am pretty sure the image size issue will solve your problem, 
but you may also get improved results using images processed in a 
slightly different way. Here are possible solutions:

SOLVING THE IMAGE SIZE ISSUE

Depending on the exact pixel count of your hdr, you need to make slight 
adjustments to one of your image angle values. Using a camera and lens 
similar to yours (Nikon D90 & AF-S Nikkor 35mm 1:1.8G), I collected a 
series of raw and .jpg processed images. In my 3 experiments, I got the 
following image sizes (shown with Matlab code):

nef = [4310 2868];
jp1= [4288 2848];
jp2= [1936 1286];

According to your notes, you set vv and vh as follows:

vv = 25.74
vh = 37.4

Now, I will recompute vv according to the pixel values and vh noted 
above, which will show slightly different vv values:

vv_t1 = 2*atand(nef(2)/nef(1)*tand(vh/2))
vv_t2 = 2*atand(jp1(2)/jp1(1)*tand(vh/2))
vv_t3 = 2*atand(jp2(2)/jp2(1)*tand(vh/2))

Notice that the answers are slightly different than vv above:

vv_t1 = 25.3864
vv_t2 = 25.3403
vv_t3 = 25.3431

A detailed explanation is available in this posting from May 16, 1994: 
http://radsite.lbl.gov/radiance/digests_html/v2n7.html#VIEW_ANGLES

SOLVING POSSIBLE METADATA ISSUE

I am not sure if this will help, but I suspect another culprit may be 
either missing metadata in your Photoshop .jpg images, or (maybe but not 
likely) the .hdr algorithm in Photoshop. I've tested the following 3 
possible solutions using the same D90 and hdrgen wrapper variants--all 
of which seem to preserve the metadata in the images, and produce 
photometrically accurate results (list members, please correct me if I 
am wrong about the above metadata assumptions)...

Method 1)

To process .nef images, see Greg's suggestion posted on February 20, 
2012: 
http://www.radiance-online.org/pipermail/hdri/2012-February/000363.html. 
This seems to be the best raw to hdr method, as it seems to produce the 
most photometrically accurate images.

My line of code looks like this: raw2hdr -o test01.hdr *.NEF (although I 
should have included a calibration image).

Or, to use .jpg images, which are less accurate photometrically (but 
still works), try either of the following procedures:

Method 2)

With .nef images still in the Nikon D90, use on-board post processing of 
.nefs to .jpgs. That should give you the right metadata embedded in the 
.jpgs, which may have been lost in the Photoshop .nef to .jpg conversion 
process. I've tried this method, and it works. On the D90: Press 
play-->select image-->press "ok"-->NEF (RAW) Processing-->(Chose your 
favorite settings)-->EXE. Repeat that process for each image. Then 
create the .hdr using hdrgen, Photosphere, or the online tool. I used 
Photosphere in this case.

Method 3)

You can do batch processing (.nef->.jpg), using Nikon's ViewNX 2 on your 
computer, as follows: Select your images, then: File-->Convert 
Files-->Convert. I've also verified that images from this method work 
with Photosphere. Or create .hdr with Photosphere, hdrgen, or the online 
tool.

Note that with methods 2 or 3, you will have .jpg images that will be 
processed using hdrgen or photosphere (or maybe Photoshop). Keep in mind 
that Method 1 is the most difficult to set up, but will produce the best 
(i.e., most photometrically accurate) results.

I hope these suggestions help (and don't lead you astray).
-Chris


On 4/23/13 10:31 PM, Avinash Gautam wrote:

_______________________________________________
HDRI mailing list
HDRI-mVArnbgP/ra1OXqM+dc5E0B+6BGkLq7r< at >public.gmane.org
http://www.radiance-online.org/mailman/listinfo/hdri

unknown view type

Avinash Gautam — 2013-04-24T03:31:24

I am trying to generate a hdri image from my camera Nikon D90 using various lenses. My Nikon camera generates a .NEF file in RAW version however I was able to generate a .HDR file using .JPG of the same images in Photoshop. After I generated the .HDR file I tried running evalglare on it which gives me an error. I understand I need to have a fish eye lens to get the full view in order to get the correct results but the purpose of my experiment is to test the error in dgp using different lenses that are not fish eye lens. I am using the -vth command and have entered the -vv and -vh as calculated for the 35mm lens on a DX format camera (25.74degrees and 37.4degrees). This gives me an error pict_update_view: unknown view type and error: invalid view specified. Can anyone please let me know where I am going wrong.

Thanks,
Avinash



_______________________________________________
HDRI mailing list
HDRI-mVArnbgP/ra1OXqM+dc5E0B+6BGkLq7r< at >public.gmane.org
http://www.radiance-online.org/mailman/listinfo/hdri

Re: [Radiance-general] Photosphere and anyhere.com?

Greg Ward — 2013-04-12T00:02:21

Anyhere.com is back online.  I haven't heard what caused the outage, but they seem to have taken care of it.  Let me know if you experience further problems.

-Greg

Re: [Radiance-general] Photosphere and anyhere.com?

Greg Ward — 2013-04-11T15:14:10

Hi Mike,

Anyhere runs on a low-budget server that occasionally experiences problem.  In this case, something more serious is going on and I just wrote to the administrator (/ business owner) to find out what.  Hopefully, we'll have it sorted out in a day or two.  Meanwhile, I will e-mail you a copy of Photosphere.

Cheers,
-Greg

P.S.  This probably belongs on the HDRI mailing list, so I'll cross-post my reply there.  Any follow-up should happen on HDRI.

>