remove light probe image from background

Hello,

(I apologize if this email went out twice ... I forgot to update my subscription to my current email...)

I am illuminating a scene with a light probe.
In the resulting rendered image I do not want to have the probe image in the background.

Does anyone know if there is a setting in Radiance that would make this possible?

Thanks for your help!

Kind Regards,

katja

Katja Doerschner
University of Minnesota
Psychology Department
218 Elliott Hall
75 East River Road
Minneapolis, MN 55455-0344
Phone:+1 612 626-8551

Quick ideas:

You could use the clipping parameters of the view definition
('-va' in particular) or the recently reported stencil method.

Thomas

···

On 19 Apr 2007, at 20:43, Katja Doerschner wrote:

I am illuminating a scene with a light probe.
In the resulting rendered image I do not want to have the probe image in the background.

Does anyone know if there is a setting in Radiance that would make this possible?

I've done it by creating a duplicate scene (your rif file) and subsititute a
overcast gensky (just something that will make the sky white, change all
your materials to black with no specular. Then render with -ab 0 - to get a
black and white "mask" Render will not take long with no bounces.
I've then converted the mask pic to a standard ldr tiff - use it in
photoshop as a masking layer.
You could also do it with the original hdrs through pcomb - you'd have to
check the manual on that.

Rob F

···

_____

From: Katja Doerschner [mailto:[email protected]]
Sent: Thursday, April 19, 2007 12:43 PM
To: [email protected]
Subject: [Radiance-general] remove light probe image from background

Hello,

(I apologize if this email went out twice ... I forgot to update my
subscription to my current email...)

I am illuminating a scene with a light probe.
In the resulting rendered image I do not want to have the probe image in the
background.

Does anyone know if there is a setting in Radiance that would make this
possible?

Thanks for your help!

Kind Regards,

katja

Katja Doerschner
University of Minnesota
Psychology Department
218 Elliott Hall
75 East River Road
Minneapolis, MN 55455-0344
Phone:+1 612 626-8551

Thanks Thomas!
- va does the trick.
katja

You could use the clipping parameters of the view definition
('-va' in particular) or the recently reported stencil method.

Katja Doerschner
University of Minnesota
Psychology Department
218 Elliott Hall
75 East River Road
Minneapolis, MN 55455-0344
Phone:+1 612 626-8551
http://vision.psych.umn.edu/~doerschner

···

On Apr 19, 2007, at 2:49 PM, Thomas Bleicher wrote:

Hi Rob,
Thank you for your suggestion.
I tried something like that before, and encountered an interesting artifact (well maybe not so interesting).
I have a perfectly specular object illuminated by a light probe. In a second image I used the same shape but made the object glow (material) and the environment black - This created something like a mask. Then I wrote a small program which picks all the non-black pixels in the mask and replaces them with the content from the first (specular) image.
To my surprise these shapes never matched up perfectly, the mask created with the non-specular material was always a few pixels bigger than the specular shape, as a result there were always a few background pixels selected. I tried other materials in combination with different lights(ambient etc) - always encountering the same problem.

katja

I've done it by creating a duplicate scene (your rif file) and subsititute a overcast gensky (just something that will make the sky white, change all your materials to black with no specular. Then render with -ab 0 - to get a black and white "mask" Render will not take long with no bounces.
I've then converted the mask pic to a standard ldr tiff - use it in photoshop as a masking layer.
You could also do it with the original hdrs through pcomb - you'd have to check the manual on that.

Rob F

From: Katja Doerschner [mailto:[email protected]]
Sent: Thursday, April 19, 2007 12:43 PM
To: [email protected]
Subject: [Radiance-general] remove light probe image from background

Hello,

(I apologize if this email went out twice ... I forgot to update my subscription to my current email...)

I am illuminating a scene with a light probe.
In the resulting rendered image I do not want to have the probe image in the background.

Does anyone know if there is a setting in Radiance that would make this possible?

Thanks for your help!

Kind Regards,

katja

Katja Doerschner
University of Minnesota
Psychology Department
218 Elliott Hall
75 East River Road
Minneapolis, MN 55455-0344
Phone:+1 612 626-8551

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

Katja Doerschner
University of Minnesota
Psychology Department
218 Elliott Hall
75 East River Road
Minneapolis, MN 55455-0344
Phone:+1 612 626-8551
http://vision.psych.umn.edu/~doerschner

···

On Apr 19, 2007, at 2:58 PM, Fitzsimmons, Rob wrote:

Hi Katja ---

This is very strange. I have done exactly what you describe countless times and never seen artifacts. How many pixels error do you see? Do you oversample the alpha mask to get anti-aliased edges? I use the small rim of fractional values in the downsampled alpha matte to crossfade between the rendering image and the mask, which always gives pixel-perfect cropping....

-- Roland

···

On Apr 19, 2007, at 10:14 PM, Katja Doerschner wrote:

Hi Rob,
Thank you for your suggestion.
I tried something like that before, and encountered an interesting artifact (well maybe not so interesting).
I have a perfectly specular object illuminated by a light probe. In a second image I used the same shape but made the object glow (material) and the environment black - This created something like a mask. Then I wrote a small program which picks all the non-black pixels in the mask and replaces them with the content from the first (specular) image.
To my surprise these shapes never matched up perfectly, the mask created with the non-specular material was always a few pixels bigger than the specular shape, as a result there were always a few background pixels selected. I tried other materials in combination with different lights(ambient etc) - always encountering the same problem.

katja

On Apr 19, 2007, at 2:58 PM, Fitzsimmons, Rob wrote:

I've done it by creating a duplicate scene (your rif file) and subsititute a overcast gensky (just something that will make the sky white, change all your materials to black with no specular. Then render with -ab 0 - to get a black and white "mask" Render will not take long with no bounces.
I've then converted the mask pic to a standard ldr tiff - use it in photoshop as a masking layer.
You could also do it with the original hdrs through pcomb - you'd have to check the manual on that.

Rob F

From: Katja Doerschner [mailto:[email protected]]
Sent: Thursday, April 19, 2007 12:43 PM
To: [email protected]
Subject: [Radiance-general] remove light probe image from background

Hello,

(I apologize if this email went out twice ... I forgot to update my subscription to my current email...)

I am illuminating a scene with a light probe.
In the resulting rendered image I do not want to have the probe image in the background.

Does anyone know if there is a setting in Radiance that would make this possible?

Thanks for your help!

Kind Regards,

katja

Katja Doerschner
University of Minnesota
Psychology Department
218 Elliott Hall
75 East River Road
Minneapolis, MN 55455-0344
Phone:+1 612 626-8551

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

Katja Doerschner
University of Minnesota
Psychology Department
218 Elliott Hall
75 East River Road
Minneapolis, MN 55455-0344
Phone:+1 612 626-8551
http://vision.psych.umn.edu/~doerschner

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

Hi Roland,
Thanks for your response!
The error is about 2-3 pixels (maybe less) all around the object. I do oversample, but I first combine the oversampled images (since I assumed the shapes would be more or less identical) and then downsample the result - this is the opposite order of your approach. Just to check, I shall try it your way + crossfading.

I am quite sure I didn't make a mistake somewhere along the image+masks generation & combination pipeline, but given that you never encountered this artifact before makes me doubt that now ...

Viele Gruesse,
katja

Roland William Fleming wrote:

···

Hi Katja ---

This is very strange. I have done exactly what you describe countless times and never seen artifacts. How many pixels error do you see? Do you oversample the alpha mask to get anti-aliased edges? I use the small rim of fractional values in the downsampled alpha matte to crossfade between the rendering image and the mask, which always gives pixel-perfect cropping....

-- Roland

On Apr 19, 2007, at 10:14 PM, Katja Doerschner wrote:

    Hi Rob,
    Thank you for your suggestion.
    I tried something like that before, and encountered an interesting
    artifact (well maybe not so interesting).
    I have a perfectly specular object illuminated by a light probe.
    In a second image I used the same shape but made the object glow
    (material) and the environment black - This created something like
    a mask. Then I wrote a small program which picks all the non-black
    pixels in the mask and replaces them with the content from the
    first (specular) image.
    To my surprise these shapes never matched up perfectly, the mask
    created with the non-specular material was always a few pixels
    bigger than the specular shape, as a result there were always a
    few background pixels selected. I tried other materials in
    combination with different lights(ambient etc) - always
    encountering the same problem.

    katja

    On Apr 19, 2007, at 2:58 PM, Fitzsimmons, Rob wrote:

        I've done it by creating a duplicate scene (your rif file) and
        subsititute a overcast gensky (just something that will make
        the sky white, change all your materials to black with no
        specular. Then render with -ab 0 - to get a black and white
        "mask" Render will not take long with no bounces.
        I've then converted the mask pic to a standard ldr tiff - use
        it in photoshop as a masking layer.
        You could also do it with the original hdrs through pcomb -
        you'd have to check the manual on that.
                 Rob F

        *From:* Katja Doerschner [mailto:[email protected]]
        *Sent:* Thursday, April 19, 2007 12:43 PM
        *To:* [email protected]
        *Subject:* [Radiance-general] remove light probe image from
        background

        Hello,

        (I apologize if this email went out twice ... I forgot to
        update my subscription to my current email...)

        I am illuminating a scene with a light probe.
        In the resulting rendered image I do not want to have the
        probe image in the background.

        Does anyone know if there is a setting in Radiance that would
        make this possible?

        Thanks for your help!

        Kind Regards,
                 katja

        Katja Doerschner
        University of Minnesota
        Psychology Department
        218 Elliott Hall
        75 East River Road
        Minneapolis, MN 55455-0344
        Phone:+1 612 626-8551

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

    Katja Doerschner
    University of Minnesota
    Psychology Department
    218 Elliott Hall
    75 East River Road
    Minneapolis, MN 55455-0344
    Phone:+1 612 626-8551
    http://vision.psych.umn.edu/~doerschner

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

------------------------------------------------------------------------

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

Well, for most purposes setting the clipping volume is a fine alternative, so it probably isn't worth worrying about this too much more. But I still don't understand how it could be happening. Do you manually specify the rays yourself and render the image with rtrace?

weird ...

-- Roland

···

On Apr 20, 2007, at 2:39 PM, Katja Doerschner wrote:

Hi Roland,
Thanks for your response!
The error is about 2-3 pixels (maybe less) all around the object. I do oversample, but I first combine the oversampled images (since I assumed the shapes would be more or less identical) and then downsample the result - this is the opposite order of your approach. Just to check, I shall try it your way + crossfading.

I am quite sure I didn't make a mistake somewhere along the image+masks generation & combination pipeline, but given that you never encountered this artifact before makes me doubt that now ...

Viele Gruesse,
katja

Roland William Fleming wrote:

Hi Katja ---

This is very strange. I have done exactly what you describe countless times and never seen artifacts. How many pixels error do you see? Do you oversample the alpha mask to get anti-aliased edges? I use the small rim of fractional values in the downsampled alpha matte to crossfade between the rendering image and the mask, which always gives pixel-perfect cropping....

-- Roland

On Apr 19, 2007, at 10:14 PM, Katja Doerschner wrote:

    Hi Rob,
    Thank you for your suggestion.
    I tried something like that before, and encountered an interesting
    artifact (well maybe not so interesting).
    I have a perfectly specular object illuminated by a light probe.
    In a second image I used the same shape but made the object glow
    (material) and the environment black - This created something like
    a mask. Then I wrote a small program which picks all the non-black
    pixels in the mask and replaces them with the content from the
    first (specular) image.
    To my surprise these shapes never matched up perfectly, the mask
    created with the non-specular material was always a few pixels
    bigger than the specular shape, as a result there were always a
    few background pixels selected. I tried other materials in
    combination with different lights(ambient etc) - always
    encountering the same problem.
    katja

    On Apr 19, 2007, at 2:58 PM, Fitzsimmons, Rob wrote:

        I've done it by creating a duplicate scene (your rif file) and
        subsititute a overcast gensky (just something that will make
        the sky white, change all your materials to black with no
        specular. Then render with -ab 0 - to get a black and white
        "mask" Render will not take long with no bounces.
        I've then converted the mask pic to a standard ldr tiff - use
        it in photoshop as a masking layer.
        You could also do it with the original hdrs through pcomb -
        you'd have to check the manual on that.
                 Rob F

        *From:* Katja Doerschner [mailto:[email protected]]
        *Sent:* Thursday, April 19, 2007 12:43 PM
        *To:* [email protected]
        *Subject:* [Radiance-general] remove light probe image from
        background

        Hello,

        (I apologize if this email went out twice ... I forgot to
        update my subscription to my current email...)

        I am illuminating a scene with a light probe.
        In the resulting rendered image I do not want to have the
        probe image in the background.

        Does anyone know if there is a setting in Radiance that would
        make this possible?

        Thanks for your help!

        Kind Regards,
                 katja

        Katja Doerschner
        University of Minnesota
        Psychology Department
        218 Elliott Hall
        75 East River Road
        Minneapolis, MN 55455-0344
        Phone:+1 612 626-8551

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

    Katja Doerschner
    University of Minnesota
    Psychology Department
    218 Elliott Hall
    75 East River Road
    Minneapolis, MN 55455-0344
    Phone:+1 612 626-8551
    http://vision.psych.umn.edu/~doerschner

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

------------------------------------------------------------------------

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

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