Classification of Infrared Image Distortions using Contrast-based Features

In this research, the spatial local mean subtraction contrast normalized coefficients are used for the classification of the infrared image distortion type. These coefficients and their products are calculated through four spatial orientations, from which 18 features are derived. A method for reduction dimension based on scatter matrices is applied in order to take a smaller number of coordinates. Hierarchical linear classifier is designed for image type distortions classification. The obtained results show that using a method for feature reduction provides overall probability slightly smaller than the overall probability of correct classification when the original 18 features are used.