luminance ratio for shadow removal

In this 开发者_开发问答paper: Human Body Tracking by Adaptive Background Models and Mean-Shift Analysis Section 3, the authors mention about luminance ratio thresholds for brightness and darkness (r1, r2). Can anyone suggest what should the values be?

---

You have them explained in this paper, sections 4.2 and 4.3.

I usually prefer to post self-contained answers, but this topic is not easy to understand without reading the whole paper.