A New Local Descriptor Based on Strings for Face Recognition

Algorithm 3

Extraction of the feature vector.

Input: Train an image (matrix)
Output: Feature vector (histogram)
	Step 1: Divide the face image into subregions
	Step 2: Extract the strings from each pixel for each region.
	Step 3: Replace all strings in each subregion with the words appearing in regional dictionary based on the Levenshtein Distance, and that for all images.
	Step 4: Construct a histogram of words for each image.