Two-Dimensional Whitening Reconstruction

In this post, I will try to implement 2D Image Whitening Reconstruction algorithm, which is basically a variant of PCA whitening, in Python programming language and of course our favorite matrix processing library NumPy.

The algorithm is described in [1] and used as a preprocessing step for facial recognition in [2].

Target Image

Image from Yale Face Dataset

Code

Resulting Image and Histograms

Let’s apply above whitening process to our target image.

We can visualize both the original and resulting histograms to observe the effect of the algorithm.