Real-Time Algorithms of Object Detection Using Classifiers

Object detection, or more generally pattern detection and recognition, can be based on many different principles. The objects can be described through their structure, shape, color, texture, etc. [Blaschko & Lampert (2009); Chen et al. (2004); Fidler & Leonardis (2007); Leibe et al. (2008); Lowe (1999); Serre et al. (2005); Viola & Jones (2001)]; therefore, a variety of object detection mechanisms was developed over time. One of the modern approaches to object detection is similarity-based detection where the objects of interest are defined through a set of examples and typically also through a set of counter-examples and the decision whether an object is an object of interest is done through machine learning-based functional block – classifier. The object detection in an image is performed by the application of the classifier on sub-windows of the image.