[Radiance-general] help pfilt and low values dgp
J. Alstan Jakubiec
alstan at jakubiec.net
Wed Aug 12 08:05:08 PDT 2015
Hi Jan,
Apologies for the -1 comment -- I didn't know it did anything special
beyond only returning DGP.
As for the version of evalglare, I was using v1.10 -- Doh! I installed
Daysim last week to get the source for Linux, and it put an old version
in my path. The 0 result issue doesn't happen with version 1.11.
Probably the Daysim installer should be updated at some point.
> /Also you should make sure, that you use a view type which really
corresponds to your lens. If your real lens is a hemispherical fish
eye and you provide -vta as view string, the angles and solid
angles are calculated wrong in evalglare and you get wrong results.
So make sure that you use -vta only if your lens is a angular fish eye./
I can also second this. I ended up with an equi-solid fisheye lens, and
there is no small amount of effort converting the images into a -vta
type for Radiance use.
Best,
Alstan
On 8/12/2015 5:39 PM, Jan Wienold wrote:
> Hi Jasper,
> hi Alstan,
>
> I guess there are several issues in that case. First of all, as Alstan
> wrote, you should crop the image and provide the correct lens
> specification to evalglare. Be aware that editing the header could be
> dangerous, because sometimes editors add strange characters or you
> don't see tabs in the header with marks the view string there as
> "invalid" . In case the header is interpreted "wrong" in evalglare,
> the results could be really random and could differ for more than 100%
> from the right ones (e.g. the calculation of the vertical illuminance,
> see presentation on the Radiance workshop in 2012). The header
> treatment is much more robust since the evalglare version 1.08, but
> still the user should take care of providing a correct header.
> evalglare is relying on a correct radiance header. To be on the safe
> side, you should always use the command option to provide the correct
> view option to evalglare.
> Also you should make sure, that you use a view type which really
> corresponds to your lens. If your real lens is a hemispherical fish
> eye and you provide -vta as view string, the angles and solid angles
> are calculated wrong in evalglare and you get wrong results. So make
> sure that you use -vta only if your lens is a angular fish eye.
> And please don't use the -1 option in evalglare!!! This is a special
> (undocumented) option to get fast results on images calculated without
> ambient calculations. If you use it for normal images, the glare
> sources might not detected correctly and you could get big
> differences. It is working mostly properly in images with large black
> areas (-ab 0).
> Since the -1 option is not robust for normal images, this option is
> undocumented. There exist also other undocumented options since 2009
> for hdr treatment(e.g. pixel overflow correction, image fillup when
> ccd-array is smaller than the projected image...), but these options
> are by purpose undocumented because they should be used only in
> special cases and can cause wrong results when not used properly.
>
> If I look at your image, I guess your calibration is not correct. The
> sun has a luminance of 2e10 cd/m2. In case of low transmittance
> glazing you still have a luminance of Xe9 cd/m2, lets say at least
> 1e9cd/m2, which is factor 200000 higher than you measured!!! And be
> aware you you might have to deal with blooming effects when you have a
> pixel overflow (especially when looking into the sun). Not sure about
> your camera setting, but if you still have an overflow for the
> shortest exposure, you should think of adding a neutral grey filter to
> reduce the overall transmittance to the ccd.
>
> Finally I can't reproduce the 0 output of evalglare. I used your image
> and used as well the 2500lux as input (even if the image is not
> cropped correctly and the view string is wrong). See here:
> /evalglare -i 2500 pmar_sin_03.hdr//
> //Notice: Low brightness scene. Vertical illuminance less than 380
> lux! dgp might underestimate glare sources//
> //dgp,dgi,ugr,vcp,cgi,Lveil: 0.311778 18.132195 21.670120 44.798004
> 27.925421 58.052605 //
>
> /But I always get an non-0 result, I never experienced this before. I
> tried the linux ,the mac and also the windows version with your image
> - it didn't happen. So which version are you using? Which operating
> system? (type evalglare -v to find out)
>
> @Alstan: Can you provide me another example where this happens as
> well? Which version are you using? Which operating system?
>
> A zero value should never appear, except your image is completely black.
>
> Best,
> Jan
>
>
>
> Am 8/11/15 um 5:33 PM schrieb J. Alstan Jakubiec:
>> Hi Jasper,
>>
>> This is one of the tricky aspects of doing glare analysis with your
>> own HDR images. A couple of pointers are below,
>>
>> * You will need to crop your image to a square aspect about the
>> image center using the pcompos
>> <http://radsite.lbl.gov/radiance/man_html/pcompos.1.html> tool.
>> There was a helpful discussion on maintaining image exposure
>> values while doing this here
>> <http://www.radiance-online.org/pipermail/radiance-general/2011-March/007701.html>.
>> * The -vv and -vh parameters are just best guesses according the
>> HDR generation software. Once you have cropped the image, you
>> will want to open the resulting HDR in a text editor and manually
>> change the header 'VIEW' field to include -vta -vv 180 -vh 180.
>> You may also specify them via the command line at this point, as
>> you have done. I like to keep it associated with the image.
>> * After that, unless I am forgetting something (others can chime
>> in), you are ready to run evalglare. I would run it with the -d
>> flag, which will report a lot of details. Most usefully, it
>> reports illuminance as derived from the image, which you can
>> compare to your measured Ev value to check the validity of the
>> HDR. If your HDR is well-calibrated, not inputting the measured
>> illuminance value should be perfectly accurate.
>> * I suspect that inputting measured illuminance is somewhat broken
>> in the current version of evalglare as I have the same problem
>> that you do. One option is to use the -1 option to evalglare,
>> which will return only a single DGP value. It seems to avoid this
>> error.
>> > evalglare -1 -i 2500 image.hdr
>>
>> By the way, to avoid some of this cropping and exposure value pain, I
>> use an image-processing tool (like PIL for Python) these days that
>> can maintain EXIF data while cropping the source jpeg files. Though
>> perhaps the cure is worse than the disease in this case..
>>
>> Best,
>> Alstan
>>
>> On 8/11/2015 11:08 PM, Jasper Overduin wrote:
>>> Dear all,
>>>
>>> I have changed my lens to one with 180 circular view to do a
>>> contrast analysis (hdrscope) and meanwhile the glare analysis in
>>> evalglare. If i use the commands in evalglare getinfo i get the (HDR
>>> composed with photosphere on a mac, calibrated with luminance
>>> pistol) i get a value for the lens -vv and -vh which is not over
>>> 100, with a lens of 180. I can imagine that the photo ratio and the
>>> lens are not the same and that causes this problem. But when I enter
>>> the external measured Ev, the value the dgp goes somehow to zero.
>>> The fact that the gdp is zero with a maximum luminance of 5600 cd/m2
>>> and Ev of 2500 lux makes me a bit suspicious. How accurate is the
>>> result of the dgp without external vertical lux? is it possible to
>>> use this value?
>>>
>>> hdr files and printscreens of evalglare
>>> https://www.dropbox.com/sh/ei1y4d2v6hapdsr/AACgVMPd1o0EdK-0q8CFvUGga?dl=0
>>>
>>>
>>> Greetings Jasper
>>>
>>>
>>> *
>>> *
>>> *Jasper Overduin*
>>> MSc. Building Technology graduate student at Delft University of
>>> Technology
>>>
>>>
>>> *S* Groenhoevelaan 3
>>> *P*2343 BP Oegstgeest
>>> *T* +31 6 15 64 48 56 (NL)
>>> *T* +56 9 51 11 76 48 (CL)
>>> *E *overduin.jasper at gmail.com
>>> *Skype *jasper.overduin
>>>
>>> On 1 May 2015 at 12:15, Jan Wienold <jan.wienold at epfl.ch
>>> <mailto:jan.wienold at epfl.ch>> wrote:
>>>
>>> Hi Jasper,
>>>
>>> I briefly looked at your image - for sure you get a low DGP
>>> value if your illuminance at camera (or eye) level is only about
>>> 300lux... It is not the matter of the fish eye lens it is a
>>> matter of your lighting condition.
>>> When I remember correctly, for the experiments I did for my PhD,
>>> the people adjusted the blinds in a way, that they had 2500-3000
>>> lux at the eye level and they were less than 20% of them
>>> dissatisfied. So a value of 300 means one order of magnitude
>>> less light at the eye level and a much lower adaptation level.
>>> So I definitely understand the low DGP value in that case. The
>>> images themselves look reasonable, so I don't think there is a
>>> problem in calibration/processing so far (at least not for these
>>> low luminance levels-it might be more tricky to calibrate for
>>> the high luminance values when you get stray-light from the
>>> multiple lenses).
>>>
>>> If all your images are like that it means you have a very low
>>> daylight contribution at the place you measure. I'm not sure if
>>> DGP is then the right way to measure glare in that case - as I
>>> wrote it is made more for the daylight oriented workplace with
>>> higher levels and also to take into account very high luminances
>>> (e.g. sun or specular reflections of the sun). DGP might be
>>> modified in future, but these experiments are just starting.
>>>
>>> Jan
>>>
>>>
>>>
>>>
>>>
>>>
>>> Am 4/30/15 um 10:41 PM schrieb Jasper Overduin:
>>>>
>>>> Thank you for the fast reply. We are still having some
>>>> problems with the outcome of Evalglare. With an external
>>>> luxometre we have done some tests now. The DGP is still
>>>> very low or zero. It seems that in almost all the case the
>>>> DGP is low. In literature we read that values above 20% are
>>>> normal. What do you think? is the data much better if we
>>>> use a full 180 degree lens?
>>>>
>>>> .hdr file https://www.dropbox.com/s/l10x7tri49btr8r/ff.hdr?dl=0
>>>> print screen cmd
>>>> https://www.dropbox.com/s/yv1lamhiirjhvxq/imp%20pan.jpg?dl=0
>>>> test.pic https://www.dropbox.com/s/nspdu01477mchsz/test.pic?dl=0
>>>>
>>>> Greetings Jasper
>>>>
>>>>
>>>>
>>>> *
>>>> *
>>>> *Jasper Overduin*
>>>> MSc Building Technology graduate student at Delft
>>>> University of Technology*
>>>> *
>>>>
>>>> *S* Groenhoevelaan 3
>>>> *P*2343 BP Oegstgeest
>>>> *T* +31 6 15 64 48 56 (NL)
>>>> *T* +56 9 51 11 76 48 (CL)
>>>> *E *overduin.jasper at gmail.com
>>>> *Skype *jasper.overduin
>>>>
>>>> On 30 April 2015 at 13:55, Jan Wienold
>>>> <jan.wienold at epfl.ch> wrote:
>>>>
>>>> Hi Jasper,
>>>>
>>>> why are you using -vth -vh 140 -vv 80 when in your
>>>> header of the HDR image the view is specified as -vtv
>>>> -vh 98.797409 -vv 75.402067 ?
>>>> Manipulating the lense type is really dangerous - in
>>>> that case you change from a perspective view to a
>>>> hemispherical fish eye view, without changing the image!!
>>>>
>>>> If I apply evalglare for your image I get 0.17 as DGP
>>>> (which is still very low, but you have only 2000cd/m2
>>>> as maximum value, so this can be expected). Be aware,
>>>> that DGP accounts only for glare from a high amount of
>>>> daylight and/or spots of extreme luminances
>>>> (>50000cd/m2), but not for contrasted glare between
>>>> task (e.g. Monitor) and immediate surroundings for
>>>> lower adaptation levels. This is subject of current
>>>> research (also here at EPFL) and there might be an
>>>> extension of the DGP in future, depending on the
>>>> outcome of new experiments.
>>>>
>>>> Back to the lens-type:
>>>> It is extremely important, that the right view type is
>>>> given to evalglare, otherwise ALL calculated values (it
>>>> doesn't matter if this is evalglare or findglare) are
>>>> wrong. These errors could be huge, more than 100% for
>>>> calculating the illuminance out of a 180 degree image.
>>>>
>>>> If you manipulate an image by pcomb, in general the
>>>> view is marked as "invalid" in the header, because with
>>>> that tool you could manipulate the image in a way, that
>>>> the original view is not valid any more. This is why
>>>> from evalglare version 1.0 on a check on the header was
>>>> included, because many people were creating wrong
>>>> headers without knowing it and then evalglare was
>>>> calculating wrong values, when the header was invalid.
>>>>
>>>> In addition for calculating the DGP it is important to
>>>> have the illuminance at camera level. evalglare
>>>> calculates this value out of the image. But if the
>>>> image does not cover 180 degree, then the calculated
>>>> value for the illuminance is too low. For that reason,
>>>> the -i option was included, so you can provide the
>>>> illuminance to evalglare (when you measure it with an
>>>> illuminance sensor).
>>>>
>>>> So in your case, you should measure the illuminance
>>>> just besides the lens.
>>>> Then (in case this is the right lens description) you
>>>> should use
>>>> evalglare -i LUXVALUE -vtv -vh 98.797409 -vv 75.402067
>>>> IMAGE_NAME
>>>> or better, if your task is always at the same place:
>>>> evalglare -i LUXVALUE -vtv -vh 98.797409 -vv 75.402067
>>>> -T 395 230 .6 -c CHECK_FILE_PICTURE IMAGE_NAME
>>>>
>>>> good luck!
>>>>
>>>> Jan
>>>>
>>>>
>>>>
>>>> Am 4/30/15 um 6:04 PM schrieb Jasper Overduin:
>>>>> Somehow cant use the command pfilt or change the
>>>>> pcomb, does this has to do with the program Radiance?
>>>>> I have installed the version of windows from the site,
>>>>> with evalglare v1.11windows . The problem is that I
>>>>> have composed a .hdr (out of 7 jpg on a mac) and after
>>>>> using the command c:/HDRI>evalglare -vth -vh 140 -vv
>>>>> 80 image.hdr all the dgp results are really low, less
>>>>> than 5%. The problem can be in the .hdr (calibrated as
>>>>> well) or is in the way evalglare is not working as it
>>>>> shoot on my computer.
>>>>>
>>>>>
>>>>> *.hdr file* (post-it is calibration point 167,98)
>>>>> https://www.dropbox.com/s/4z69y358yt8ii4z/1_sv_am.hdr?dl=0
>>>>> *images* original
>>>>> https://www.dropbox.com/sh/adyyxjkv6eykyek/AAC96QUTpLh_Ef2U8Mppki8ta?dl=0
>>>>> *command printscreen*
>>>>> https://www.dropbox.com/s/0pbn105p0z0iinu/Imp%20pan.jpg?dl=0
>>>>>
>>>>>
>>>>> Need the help!
>>>>>
>>>>> Greetings Jacobus
>>>>>
>>>>> *
>>>>> *
>>>>> *Jasper Overduin*
>>>>> MSc Building Technology graduate student at Delft
>>>>> University of Technology*
>>>>> *
>>>>>
>>>>> *S* Groenhoevelaan 3
>>>>> *P*2343 BP Oegstgeest
>>>>> *T* +31 6 15 64 48 56 (NL)
>>>>> *T* +56 9 51 11 76 48 (CL)
>>>>> *E *overduin.jasper at gmail.com
>>>>> *Skype *jasper.overduin
>>>>>
>>>>>
>>>>> _______________________________________________
>>>>> Radiance-general mailing list
>>>>> Radiance-general at radiance-online.org <mailto:Radiance-general at radiance-online.org>
>>>>> http://www.radiance-online.org/mailman/listinfo/radiance-general
>>>>
>>>> --
>>>> Dr.-Ing. Jan Wienold
>>>> Ecole Polytechnique Fédérale de Lausanne (EPFL)
>>>> EPFL ENAC IA LIPID
>>>>
>>>> http://people.epfl.ch/jan.wienold
>>>> LE 1 111 (Office)
>>>> Phone +41 21 69 30849
>>>>
>>>>
>>>> _______________________________________________
>>>> Radiance-general mailing list
>>>> Radiance-general at radiance-online.org
>>>> <mailto:Radiance-general at radiance-online.org>
>>>> http://www.radiance-online.org/mailman/listinfo/radiance-general
>>>>
>>>>
>>>>
>>>>
>>>>
>>>> _______________________________________________
>>>> Radiance-general mailing list
>>>> Radiance-general at radiance-online.org <mailto:Radiance-general at radiance-online.org>
>>>> http://www.radiance-online.org/mailman/listinfo/radiance-general
>>>
>>> --
>>> Dr.-Ing. Jan Wienold
>>> Ecole Polytechnique Fédérale de Lausanne (EPFL)
>>> EPFL ENAC IA LIPID
>>>
>>> http://people.epfl.ch/jan.wienold
>>> LE 1 111 (Office)
>>> Phone +41 21 69 30849
>>>
>>>
>>> _______________________________________________
>>> Radiance-general mailing list
>>> Radiance-general at radiance-online.org
>>> <mailto:Radiance-general at radiance-online.org>
>>> http://www.radiance-online.org/mailman/listinfo/radiance-general
>>>
>>>
>>>
>>>
>>> _______________________________________________
>>> Radiance-general mailing list
>>> Radiance-general at radiance-online.org
>>> http://www.radiance-online.org/mailman/listinfo/radiance-general
>>
>>
>>
>> _______________________________________________
>> Radiance-general mailing list
>> Radiance-general at radiance-online.org
>> http://www.radiance-online.org/mailman/listinfo/radiance-general
>
> --
> Dr.-Ing. Jan Wienold
> Ecole Polytechnique Fédérale de Lausanne (EPFL)
> EPFL ENAC IA LIPID
>
> http://people.epfl.ch/jan.wienold
> LE 1 111 (Office)
> Phone +41 21 69 30849
>
>
> _______________________________________________
> Radiance-general mailing list
> Radiance-general at radiance-online.org
> http://www.radiance-online.org/mailman/listinfo/radiance-general
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.radiance-online.org/pipermail/radiance-general/attachments/20150812/44008c7a/attachment-0001.html>
More information about the Radiance-general
mailing list