Deprecated: preg_replace(): The /e modifier is deprecated, use preg_replace_callback instead in /www/htdocs/w0068b6d/smf/Sources/Load.php(225) : runtime-created function on line 3
a random report about the beta
forum@fdrtools.com
June 20, 2019, 09:07:28 AM *
Welcome, Guest. Please login or register.

Login with username, password and session length
News: SMF - Just Installed!
 
   Home   Help Search Login Register  
Pages: [1] 2
  Print  
Author Topic: a random report about the beta  (Read 20801 times)
Carel
Resident
***
Posts: 51


View Profile
« on: June 23, 2006, 01:38:50 AM »

The histogram *is* showing on my PC and I like the interface for adjusting black & white points much better.

Please look at jpgs I will post http://www.sphericalpanoramas.com/zips.html
shortly. It will be called "fringing and color noise".
I also have included jpgs of the "other" tonemapping program. It also shows some color fringing (the blue between mountains and clouds), but less so.
Unique for the fdrbeta was the brightly colored specks that show amongst the roots of the oak-brush. It does not show in fdrtools v 1.6 and it does not show in the "other" program. It only showed up in the FDRToolsbeta.
I used the same compression for v1.6 and beta. I did not include a jpg of v1.6, because there is nothing to see, except stronger banding between the greys of the clouds. I tried increasing the ETR, but it did not have much effect on the color fringing. The finished panorama, done with the "other" tonemapping operator, can be seen here:
http://sphericalpanoramas.com/sgm_flowers.html
Logged
AndreasSchoemann
Administrator
Resident
*****
Posts: 551


View Profile Email
« Reply #1 on: June 23, 2006, 01:37:40 PM »

Quote from: "Carel"
The histogram *is* showing on my PC

Good to know.

Quote
Please look at jpgs I will post http://www.sphericalpanoramas.com/zips.html
shortly. It will be called "fringing and color noise".
I also have included jpgs of the "other" tonemapping program. It also shows some color fringing (the blue between mountains and clouds), but less so.
Unique for the fdrbeta was the brightly colored specks that show amongst the roots of the oak-brush. It does not show in fdrtools v 1.6 and it does not show in the "other" program. It only showed up in the FDRToolsbeta.
I used the same compression for v1.6 and beta. I did not include a jpg of v1.6, because there is nothing to see, except stronger banding between the greys of the clouds. I tried increasing the ETR, but it did not have much effect on the color fringing. The finished panorama, done with the "other" tonemapping operator, can be seen here:
http://sphericalpanoramas.com/sgm_flowers.html

Thank you for your example images, I'll have a look at it this evening.
Logged
AndreasSchoemann
Administrator
Resident
*****
Posts: 551


View Profile Email
« Reply #2 on: June 24, 2006, 07:47:33 AM »

Hi Carel,

I've had a look upon the "egyptian" example and found a color banding around the lamps that looks not ok and is not there in 1.6.

In the "color noise and fringing" images I can see the color banding and the speckles. I would like to investigate what is going wrong there but need the original images for that. Could you make them available to me, please?
Logged
Carel
Resident
***
Posts: 51


View Profile
« Reply #3 on: June 24, 2006, 06:34:42 PM »

Hi Andreas,

Quote
I've had a look upon the "egyptian" example and found a color banding around the lamps that looks not ok and is not there in 1.6.


I did not try the Egyptian myself on the beta. That was left on the page for another problem. But is has a tricky color banding area, so is probably a good test pano.

Do the originals for "speckle inspection" need to be at at full 4000x2000 or should I reduce to f.i. 2000x1000 ? I guess resizing may give totally different results because of local contrast ?
Logged
AndreasSchoemann
Administrator
Resident
*****
Posts: 551


View Profile Email
« Reply #4 on: June 24, 2006, 07:06:31 PM »

Quote
Do the originals for "speckle inspection" need to be at at full 4000x2000 or should I reduce to f.i. 2000x1000 ? I guess resizing may give totally different results because of local contrast ?

If the downsized version shows the same artefacts then it is of course ok.
Logged
Carel
Resident
***
Posts: 51


View Profile
« Reply #5 on: June 24, 2006, 08:53:58 PM »

I have tried both reducing to 2000x1000 and cropping, but the problem only shows in the full sized equirects coming out of PTGui. I did not check if a simple re-save in Photoshop would make a difference...

I uploaded the three full-sized bracketed equirects. They are in one rar file at http://www.sphericalpanoramas.com/zips.html
Logged
AndreasSchoemann
Administrator
Resident
*****
Posts: 551


View Profile Email
« Reply #6 on: June 27, 2006, 07:46:41 PM »

Quote from: "Carel"
I have tried both reducing to 2000x1000 and cropping, but the problem only shows in the full sized equirects coming out of PTGui. I did not check if a simple re-save in Photoshop would make a difference...

Hi Carel,

I have an intermediate report for you.

The problem with the speckles is solved now. I will publish a second beta on Saturday that includes this fix and a bunch of other fixes as well. I would like to give it to you earlier but I'm currently on the way with my camera and iBook and have no access to my Windows development environment. So I've fixed the issues on OS X so far only...

I've had a look for the "color fringes" problem as well. I'm now pretty sure that there is a problem in your image editing workflow before the HDR generation. If you look at the original images you will find lots of pixels in the leaves and branches of the oak that are color corrupted, having a reddish tint. This is emphasized by tone mapping and makes the red fringing. The original images show other artefacts as well: banding in combination with color shifts in the clouds and the mountain tops - these are emphasized by the tone mapping as well.

So far I've investigated your "fringing and color noise" example. The "egyptian" example is still to be done in detail.
Logged
Carel
Resident
***
Posts: 51


View Profile
« Reply #7 on: June 29, 2006, 08:11:06 AM »

I dont know what is wrong with my workflow. I notice strange things too in some of the images, but I just follow Erik Krause's workflow for using ARC as discribed Here: http://www.panotools.info/mediawiki/index.php?title=RAW_dynamic_range_extraction
When I use dcraw there are no fringes, but one can not do CA and correct for vignetting in dcraw. And CA after demosaicing does not remove all the CA.

I never experienced that much of a problem with the egyptian theatre. That was on the zips page for another reason. It has some fringing around the spotlights on the side-walls, but I always thought that was to be expected under those conditions. It seems that the new version of the Bibble raw converter can do CA and vignetting. So I will try that out.
Logged
Carel
Resident
***
Posts: 51


View Profile
« Reply #8 on: June 29, 2006, 06:01:49 PM »

On further inspection of the raw images, I see strange colored areas on f.i. walls. So a white indoor wall will have large areas that have pink patches in areas that have a slightly different light intensity. At first I thought this was caused during de-mosaicing, but I also see this in the raw image. Is this a limitation of the image chip in the camera?

Andreas, could you please look at the link I posted yesterday to the panotools wiki, about maximizing the dynamic range during raw to tif conversion? I would like to figure out what I am doing wrong.

Or is this a limitation of the hardware that one runs into faster when shooting panoramas, where one always has huge differences in light intensity?

Carel
Logged
AndreasSchoemann
Administrator
Resident
*****
Posts: 551


View Profile Email
« Reply #9 on: June 29, 2006, 08:03:37 PM »

Quote from: "Carel"
On further inspection of the raw images, I see strange colored areas on f.i.

Eh, what does "f.i." mean? Haven't read that before...

Quote
walls. So a white indoor wall will have large areas that have pink patches in areas that have a slightly different light intensity. At first I thought this was caused during de-mosaicing, but I also see this in the raw image. Is this a limitation of the image chip in the camera?

I can't think of such a limitation. As a first step I would check if the in-camera jpegs show the same phenomenon. If so, it might be an indication of a camera problem. If not, I would then try different raw converters to find out which one causes the trouble...

Quote
Andreas, could you please look at the link I posted yesterday to the panotools wiki, about maximizing the dynamic range during raw to tif conversion? I would like to figure out what I am doing wrong.

Maybe Erik has more expertise than I have on RAW conversion with ACR. I must confess I haven't used it that often...

Quote
Or is this a limitation of the hardware that one runs into faster when shooting panoramas, where one always has huge differences in light intensity?

No I don't think so. You said that you have no problems with color fringing when you use dcraw. Well, that's a good starting point for your investigation. I'm pretty sure that your problem is related to some software tool you use, maybe it is ACR.

Unfortunately I'm too busy at the moment to go into a detailed analysis of your problem - though finding the troublemaker is surely very instructive. Your "egyptian" example and some night shots I've done during the last days resp. nights have drawn my attention to an other problem of the current beta version I'm now working on...
Logged
Carel
Resident
***
Posts: 51


View Profile
« Reply #10 on: June 30, 2006, 12:57:26 AM »

f.i. just means "for instance", maybe it is not an "official" abreviation...

Someone suggested I should check clipping channel by channel. At the moment I have only checked clipping on all channels combined. But the problem is that the bracketed image that is meant for the shadows, will almost always have clipped areas. So if those clipped areas are included in the final tonemapped image, they will produce problems..?
Logged
AndreasSchoemann
Administrator
Resident
*****
Posts: 551


View Profile Email
« Reply #11 on: June 30, 2006, 07:38:37 AM »

Quote
Someone suggested I should check clipping channel by channel. At the moment I have only checked clipping on all channels combined.

dcraw used with the "don't clip colors" option causes overexposed areas to look pink, have a look at http://fdrtools.com/faq_e.php#105.
ACR doesn't do this. Afaik they try to reconstruct clipped information using the red channel info. At any rate overexposed areas won't look pink but grey.
Logged
AndreasSchoemann
Administrator
Resident
*****
Posts: 551


View Profile Email
« Reply #12 on: July 05, 2006, 08:44:24 AM »

Carel,

I've just releases v1.7beta2. I'm sorry for the short delay but a hardware failure on the main development machine caused a few days downtime.

The "egyptian" problem is still under investigation.
Logged
Carel
Resident
***
Posts: 51


View Profile
« Reply #13 on: July 07, 2006, 04:48:34 PM »

What improvements should I look for in the new beta?

One request:
Would it be possible to reset 360degrees, zenith, nadir all at the same time?
Now one has to wait for re-processing after every step.

I was experimenting with the "belgium.hdr" from Fattal's site and noticed that FDRTools usually gives a final result with higher saturation in the exposed tif than in the preview of the navigator window. This is true for version 1.6 and the two 1.7 betas. Compared to the Navigator, the compression also "seems" higher in the exposed tif, by a bit over on point. I may be wrong on that one...
Logged
AndreasSchoemann
Administrator
Resident
*****
Posts: 551


View Profile Email
« Reply #14 on: July 07, 2006, 05:36:33 PM »

Quote
What improvements should I look for in the new beta?


Please verify that the speckles are gone now. That's it for beta2.



Quote
One request:

Would it be possible to reset 360degrees, zenith, nadir all at the same time?

Now one has to wait for re-processing after every step.


Yes. I will add a "Refresh" button as proposed in possible enhancements.



Quote
I was experimenting with the "belgium.hdr" from Fattal's site and noticed that FDRTools usually gives a final result with higher saturation in the exposed tif than in the preview of the navigator window. This is true for version 1.6 and the two 1.7 betas. Compared to the Navigator, the compression also "seems" higher in the exposed tif, by a bit over on point. I may be wrong on that one...


No, you are right. The current situation is that Navigator and final image yield more or less different results, depending mainly on the image size difference between Navigator and original. The final image will in general look more contrasty and sometimes overdone (with e.g. muddy looking surfaces), depending on the size difference and on the character of the scene. Saturation might also differ somewhat but this can easily be corrected.



Currently it is hard to predict how the final result will look like. What one can do is

- take this behaviour into consideration and use conservative settings for Compression and Contrast.

- downsize the input images (if this is an option) to reduce the "size difference effect".

Admittedly this situation is not satisfying and hence I will put all my efforts for the next version into terminating this problem.
Logged
Pages: [1] 2
  Print  
 
Jump to:  

Powered by MySQL Powered by PHP Powered by SMF 1.1.11 | SMF © 2006-2009, Simple Machines LLC Valid XHTML 1.0! Valid CSS!