7.1 Object Detection and Recognition in Computer Vision

Object detection is one kind of image segmentation based on geometrical and statistical features of objects, and object recognition refers to identification of a specific object from other objects including similar objects. Recognition is a further work that takes place after detection; in other words, detection is the important first stage of recognition. For example, face detection in a scene is to find human facial regions that have symmetrical geometrical features and specific skin colour (or other features), while face recognition is to distinguish a specific person from many other people.

7.1.1 Basic Concepts

In general, a raw input image needs to be preprocessed, such as denoising and segmentation. Denoising is mainly accomplished by different low-pass filtering in a spatial or transform domain. The conventional segmentation methods are of three types: region based, edge based and motion based. The region based segmentation includes threshold segmenting by setting one or several intensity values as the thresholds in grey or colour histograms to partition the image, or segmentation with the aid of entropy, region growing from some seed pixels, clustering, graph segmentation, region marking, region splitting and merging and so on. Edge based methods include edge extraction, active contour (snake) models and so on. The motion based methods mainly use the difference between two adjacent frames and optical flow. In the preprocessing ...

Get Selective Visual Attention: Computational Models and Applications now with the O’Reilly learning platform.

O’Reilly members experience books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.