accurate rendering of concavities

Dear fellow Radiance users,

Recently, I did an experiment that involved the study of subtly-shaded, locally-concave surface patches, including contours that were indirectly illuminated in some areas. The problem was, in these deeper regions of the object, the ambient lighting term dominated the rendered patch.

In the physical world, there is probabilistically only one darkest region (i.e., a single darkest pixel in HDRI) in any particular image of a scene. However, with Radiance, I find large patches that share the same darkest luminance value, and the subtle shading information is lost. In my last experiment, I was able to overcome this problem with a hack: I added a large, dim area light for fill. This, in effect, caused the cosine reflectance function to dominate the shading values, which is not the same as indirect lighting effects caused by inter-reflections and contour changes. Unfortunately, although that hack worked for my purposes at the time, it would not be the right approach for my current project. How does one eliminate this problem in his/her renderings using Radiance?

In general, is there a way to recreate an environment using Radiance that ensures photometric accuracy in subtly-shaded concave regions of convoluted objects (e.g., caused by gradual orientation changes, umbras, self-occlusions, and inter-reflections)?

Thank you for your thoughtful considerations.

Sincerely,
Chris Kallie

Did you try rendering first with an "overture" calculation -- create and discard a lower-resolution version of the same image in order to fill the ambient cache? This sometimes helps. Also, reducing the -aa value down as low as 0.01 should improve the subtle shading in the indirect regions.

-Greg

···

From: Chris Kallie <[email protected]>
Subject: [Radiance-general] accurate rendering of concavities
Date: July 13, 2015 7:56:55 PM PDT

Dear fellow Radiance users,

Recently, I did an experiment that involved the study of subtly-shaded, locally-concave surface patches, including contours that were indirectly illuminated in some areas. The problem was, in these deeper regions of the object, the ambient lighting term dominated the rendered patch.

In the physical world, there is probabilistically only one darkest region (i.e., a single darkest pixel in HDRI) in any particular image of a scene. However, with Radiance, I find large patches that share the same darkest luminance value, and the subtle shading information is lost. In my last experiment, I was able to overcome this problem with a hack: I added a large, dim area light for fill. This, in effect, caused the cosine reflectance function to dominate the shading values, which is not the same as indirect lighting effects caused by inter-reflections and contour changes. Unfortunately, although that hack worked for my purposes at the time, it would not be the right approach for my current project. How does one eliminate this problem in his/her renderings using Radiance?

In general, is there a way to recreate an environment using Radiance that ensures photometric accuracy in subtly-shaded concave regions of convoluted objects (e.g., caused by gradual orientation changes, umbras, self-occlusions, and inter-reflections)?

Thank you for your thoughtful considerations.

Sincerely,
Chris Kallie

Will try these two approaches.
Thanks for the suggestions!
-Chris

···

On 7/13/15 11:47 PM, Greg Ward wrote:

Did you try rendering first with an "overture" calculation -- create and discard a lower-resolution version of the same image in order to fill the ambient cache? This sometimes helps. Also, reducing the -aa value down as low as 0.01 should improve the subtle shading in the indirect regions.

-Greg

From: Chris Kallie <[email protected]>
Subject: [Radiance-general] accurate rendering of concavities
Date: July 13, 2015 7:56:55 PM PDT

Dear fellow Radiance users,

Recently, I did an experiment that involved the study of subtly-shaded, locally-concave surface patches, including contours that were indirectly illuminated in some areas. The problem was, in these deeper regions of the object, the ambient lighting term dominated the rendered patch.

In the physical world, there is probabilistically only one darkest region (i.e., a single darkest pixel in HDRI) in any particular image of a scene. However, with Radiance, I find large patches that share the same darkest luminance value, and the subtle shading information is lost. In my last experiment, I was able to overcome this problem with a hack: I added a large, dim area light for fill. This, in effect, caused the cosine reflectance function to dominate the shading values, which is not the same as indirect lighting effects caused by inter-reflections and contour changes. Unfortunately, although that hack worked for my purposes at the time, it would not be the right approach for my current project. How does one eliminate this problem in his/her renderings using Radiance?

In general, is there a way to recreate an environment using Radiance that ensures photometric accuracy in subtly-shaded concave regions of convoluted objects (e.g., caused by gradual orientation changes, umbras, self-occlusions, and inter-reflections)?

Thank you for your thoughtful considerations.

Sincerely,
Chris Kallie

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