Question about "PCA coefficient of the remapping curves"

[Alex Zou]

How exactly the remapping curve is formulated and how to derive the PCA coefficient?

[Vladimir Bychkovsky]

Luminance curve maps luminance of the "before" image to luminance of the "after" image. Each curve is represented as 51 floating point values. Combining curves from 5000 images results in 5000x51 matrix. If you run PCA in this matrix, you will get 5000x51 coefficients. Most of the paper is about learning the first of these coefficients (I.e. the one that captures most of the variance.)

Vlad.

On Mar 22, 2016, at 12:00 AM, Alex Zou <zou....@gmail.com> wrote:

How exactly the remapping curve is formulated and how to derive the PCA coefficient?

--
You received this message because you are subscribed to the Google Groups "MIT-Adobe FiveK Dataset" group.
To unsubscribe from this group and stop receiving emails from it, send an email to fivek-datase...@googlegroups.com.
To post to this group, send email to fivek-...@googlegroups.com.
Visit this group at https://groups.google.com/group/fivek-dataset.
For more options, visit https://groups.google.com/d/optout.