Basel Illumination Prior

92 views
Skip to first unread message

Yuhang Shi

unread,
May 3, 2022, 10:41:57 PM5/3/22
to scalismo-faces
Hi,

Basel Illumination Prior 2017 is a excellent work.

I want to apply SH lighting to another model. But, I don't know how to sample SH coefficients from Basel Illumination Prior. Is it a random selection directly from .rps files?

egger.b...@gmail.com

unread,
May 4, 2022, 2:10:48 AM5/4/22
to scalismo-faces
Hi,

yes that is one way to use the prior:
- randomly sample an rps file would be a sample from the empirical distribution

We however also experimented building a multivariate normal distribution and applying PCA to the SH coefficients and that also works reasonably well. The data is not perfectly Gaussian distributed, but close enough!

Best
Bernhard

Message has been deleted

Yuhang Shi

unread,
May 4, 2022, 7:28:38 AM5/4/22
to scalismo-faces
Thank you for your reply.

I compute the mean value vector and covariance matrix to build a new  multivariate normal distribution. Then, I take random samples from the  distribution as SH coefficients. But, the rendered images are dark, just like the image shows. Is it that i'm missing something?
微信截图_20220504191152.png

Bernhard Egger

unread,
May 4, 2022, 8:28:43 AM5/4/22
to Yuhang Shi, scalismo-faces
Hi,

the illumination prior comes with a jar file that provides this functionality, that calculates the multivariate normal and sample from it.
For me it looks like you might have forgotten to add the mean?

Best
Bernhard
--
0 1 0
0 0 1
1 1 1


--
You received this message because you are subscribed to the Google Groups "scalismo-faces" group.
To unsubscribe from this group and stop receiving emails from it, send an email to scalismo-face...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/scalismo-faces/059bc2bd-9a25-4e57-98a2-aca2489acad9n%40googlegroups.com.

Yuhang Shi

unread,
May 4, 2022, 9:11:25 AM5/4/22
to scalismo-faces
Thank you!! 
After sampling, I added the mean value and got much better results.

Yuhang Shi

unread,
May 5, 2022, 2:33:32 AM5/5/22
to scalismo-faces
Hello, it's me again. Sorry to bother you.

I found that the direction of illumination in most images is from bottom to top, and some of these images are very bright. 

In the paper, the illumination conditions are normalized relative to the camera position. Does this mean that the light direction is related to the camera position?If so, how to implement normalization?

thank you
在2022年5月4日星期三 UTC+8 20:28:43<egger.b...@gmail.com> 写道:

Bernhard Egger

unread,
May 5, 2022, 2:42:54 AM5/5/22
to Yuhang Shi, scalismo-faces
Hi,

we do provide a running example in the jar file which is released with the illumination prior.
Please check there if it fulfills your requirements of a natural illumination prior.
It also serves as a reference implementation where you can see all the implementation details.

There are also previous questions on this mailing list respectively in the parametric image generator github repository that might be helpful for you to check

Best
Bernhard
--
0 1 0
0 0 1
1 1 1

Reply all
Reply to author
Forward
0 new messages