Feature Detection

• Published in: Guide to Medical Image Analysis pp. 173 - 207
• Main Author: Toennies, Klaus D
• Format: Book Chapter
• Language: English
• Published: United Kingdom Springer London, Limited 2017; Springer London
• Series: Advances in Computer Vision and Pattern Recognition
• Subjects: Corner Detector; Edge Tracking; Local Binary Pattern; Shape Context; Sift Feature
• ISBN: 9781447173182; 144717318X
• ISSN: 2191-6586; 2191-6594
• DOI: 10.1007/978-1-4471-7320-5_5

Image-based features such as key point locations or potential parts of object boundaries can be extracted from local image characteristics. Boundary parts are generated from results of an edge-enhancement step, while key point locations are local extrema of some local object property. Features may also be computed from samples of an object’s boundary or interior. Potential object boundary parts are used for detecting or delineating objects in images. Key points may in some simple cases also be used to detect objects. In most cases, however, object characteristics are too complex to be captured by the attributes of a key point. They can be important attributes nonetheless. Key points define an object-dependent reference system in which they may be used to map objects of the same class onto each other. Region characteristics describe object boundaries just indirectly. They are often more robustly to compute as they often require only a rough estimate of the interior for their computation. Hence, they are more often used for object detection rather than for object delineation.