Histogram of Oriented Gradients (HOG) features descriptor were first introduced by Navneet Dalal and Bill Triggs. Their work was focused on pedestrian detection.Since then ,HOG is extensively used for object detection in computer vision field for various reasons . First ,It is easy to use with discriminate classifiers such as support vector machine. Second, HOG tries to capture shape of an object from edges(gradients) .Therefor HOG gives good results to identify object from cluttered background without using any segmentation algorithm.
Research project "Inverting and Visualizing Features for Object Detection" by MIT was very helpful to visualize how HOG features look like.There is online tool on project site where you can upload image and visualize HOG features.There is also good example of people detection using HOG in OpenCV. However , I remained puzzled , how HOG descriptor can be used with classification algorithm. And for that I required to first understand how to construct HOG descriptor. And as result of my learning I created infographics to help you understand the main idea.
In the above infographics ,I took HOG descriptor proposed by Dalal and Triggs for pedestrian detection for illustration.But you can construct HOG descriptor for any object once you understand basic idea.
P.S
Online tool to visualize HOG
"Inverting and Visualizing Features for Object Detection" - http://web.mit.edu/vondrick/ihog/
