zca whiten of an image — ZCAwhiten • OpenImageR

this function performs zca-whitening to a 2- or 3- dimensional image

Usage

ZCAwhiten(image, k, epsilon)

Arguments

image: a matrix, data frame or 3-dimensional array where the third dimension is equal to 3
k: an integer specifying the number of components to keep when svd is performed (reduced dimension representation of the data), between 1 and the number of rows of the image
epsilon: a float specifying the regularization parameter

Value

a matrix or 3-dimensional array where the third dimension is equal to 3

Details

Whitening (or sphering) is the preprocessing needed for some algorithms. If we are training on images, the raw input is redundant, since adjacent pixel values are highly correlated. When using whitening the features become less correlated and all features have the same variance.

References

http://ufldl.stanford.edu/tutorial/unsupervised/PCAWhitening/

Examples


path = system.file("tmp_images", "1.png", package = "OpenImageR")

image = readImage(path)

res = ZCAwhiten(image, k = 20, epsilon = 0.1)