CODEWITHSUNDEEP
image-processinghistogram
kcc__
kcc__

Reputation: 1648

Histogram Normalization Features

I am confused how I extract features using histogram normalization. Maybe this sounds too silly but sorry about been silly I am just confused. I have normalized the image but how i represent it as a feature. Is it the histogram becomes the feature or is it the concept of HOG (histogram of Oriented Gradient). I am confused how I represent histogram as a features. Can I get some explanation on this please.

Upvotes: 1

Views: 1352

Answers (1)

Zaphod
Zaphod

Reputation: 1927

A feature descriptor may be defined as any property of a pattern that describes it in part or in full. When you extract the histogram of an image, you're extracting a few properties of the image that may be applied or exploited in various ways. For example, if you have a majority of the histogram population biased towards the darker side, you can conclude that the image is either filled with dark patterns or is not illuminated sufficiently. So the histogram itself is an effective feature descriptor.

To make histograms comparable, normalisation is usually necessary. For example, you may want to compare images of different sizes. In this case, the overall population of the histograms will be different for the two images. But once you normalise the histograms, they become comparable, which in turn makes the feature description effective and usable.

Upvotes: 3

Related Questions