search

RockStarCoders / alienMarkovNetworks

Using MRFs and CRFs for computer vision problems.
21 stars 9 forks source link

add lbp features #30

Closed jsherrah closed 11 years ago

jsherrah commented 11 years ago

Needs to use histograms, use same strategy as HOG for now:

http://scikit-image.org/docs/dev/auto_examples/plot_local_binary_pattern.html

amb-enthusiast commented 11 years ago

The sckit-image docs for the LBP function state that it takes in an (N x M) greyscale image and returns an (N x M) LBP image: http://scikit-image.org/docs/dev/api/skimage.feature.html?highlight=lbp#skimage.feature.local_binary_pattern

The standard resizing function in FeatureGenerator.py should be fine...

amb-enthusiast commented 11 years ago

Just done a few small tests on input & output image sizes with various LBP settings, and input & output image sizes match - we get a LBP feature value for every pixel from the scikit-image LBP function.