Local binary patterns and Histogram of Gradients implementation using python3

The program is coded using Windows 10 (64 bit) operating system with Python version 3.5.2.

Local Binary Pattern

• Local Binary Pattern (LBP) is a feature descriptor used in texture analysis LBP labels the pixels of an image by thresholding the neighborhood of each pixel and considers the result as a binary number.
• For given image divide the image into n*n window (e.g. n = 8)
• For each pixel in a cell, compare the pixel to each of its 8 neighbors (on its left-top, left-middle, left-bottom, right-top, etc.). Follow the pixels along a circle, i.e. clockwise or counter-clockwise.
• Where the center pixel's value is greater than the neighbor's value, write "0". Otherwise, write "1". This gives an 8-digit binary number (which is usually converted to decimal for convenience).
• Compute the histogram, over the cell, of the frequency of each "number" occurring (i.e., each combination of which pixels are smaller and which are greater than the center). This histogram can be seen as a 256-dimensional feature vector. Optionally normalize the histogram.
• Concatenate (normalized) histograms of all cells. This gives a feature vector for the entire window.

Histogram of Gradients

• The histogram of oriented gradients (HOG) is a feature descriptor used in computer vision and image processing for the purpose of object detection.
• In HOG detects Object in the image by the distribution of intensity gradients
• Steps for calculating the HOG:
  • Computation of gradient value. This can be done by applying 1-D centered, point discrete derivative mask in one or both of the horizontal and vertical directions.
  • Filtering mask
  • [-1 , 0 , 1] and transpose of [ -1 , 0 , 1]
• Second step includes the creation of cell histograms as per the given number of bins.

Code

feature_descriptor.py

Generate Descriptors based on Model HOG and LBP

generate_HOG_descriptor

Generate Histogram of gradients of given image
:param image: Input image of Hand (1200 * 1600)
:return: feature_descriptor(9576), HOG_image(120*160)

generate_LBP_descriptor

Split the given image into 100x100 windows, compute Local Binary Pattern features for each window, and concatenate these to obtain a unified feature descriptor
:param image: Input image of Hand (1200 * 1600)
:return: feature_descriptor(49152(12*16*256)), LBP_image(1200*1

similarity.py

Similarity functions for finding similarity between two descriptors

distance_of_LBP

Return euclidean distance
:param LBP1: LBP of first image
:param LBP2: LBP of second image
:return: distance value

find_k_similar_LBP

Find k similar images of given image.
Retrieve all LBP_image_descriptors from database and compare this LBP descriptors with given image_HOG_descriptor using euclidean distance and find k similar images
:param image: Input image of Hand (1200 * 1600)
:param k: number of similar image you want to find
:return: List of ksimilarimages (imagename, distance_major(euclidian_distance))

cos_sim

Function for finding cosine-similarity
:param a: list1
:param b: list2
:return: similarity value [0:1]

euclidian_distance

Function for finding euclidean distance
:param a: list1
:param b: list2
:return: distance

Package Version Installation

The following packages are necessary to run the program:

Scikit-image 0.15.0

pip install scikit-image

Pymongo 3.8.0

pip install pymongo

Numpy 1.16.4

pip install numpy

Matplotlib 3.1.1

pip install matplotlib

Tkinter

pip install tkinter

Glob

pip install glob