Classifications of Image Features: A Survey

Abstract

Computer imaging is a complex multi-discipline science with broad application and well developed theory. A brief
knowledge of computer imaging is presented in this paper. An image feature is a descriptor of an image, which can
avoid redundant data and reduce the effects of noise and variance. In computer imaging, feature selection is vital
for researchers and processors. Feature extraction and image processing are based on the mathematical selection,
computation and manipulation of image features with high efficiency, robustness and invariance. Common image
features are expressed under definitions of feature measurements, which is stated in this paper. This paper mainly
brings an overall presentation of different sorts of image features and classifies them into specified types. Based
on different purposes of application, three main ways are put forward in this paper to categorize image features.
The first one is based on the nature of the image. The features applied to a binary image are different from the
ones applied to a gray-level image or a color image. The second classification separates visible features from
invisible features. The last one classifies image features into global image features and local image features. A
clear statement is given for each way of classification and each type of image feature. Every image feature has
both merits and defects, hence when selecting features for further image application, a clear cognition of different
features is required. Well applied image features and the algorithms related to them are highlighted in this paper
with analysis and comparison.