Using thresholding techniques for object detection in infrared images

Image processing techniques play an important role in military applications. Image binarization could be understood as a process of pixel values segmentation of grayscale image into two value groups, zero as a background and 1 as a foreground. In simple humor application of object detection we assume that contrast distribution of foreground is uniformed and without background noise or that variation in contrast does not exist. However, in complex cases previous conditions are inappropriate as variation in contrast exists and it does include background noise, etc. This paper deals with object detection in infrared images for military application using an image binarization step. Military targets are detected in different conditions such as winter condition, summer condition, at night etc. This paper focuses on combination of two methods of image binarization. One is the global binarization method proposed by Otsu and the other one is the local adaptive threshold technique. The global binarization method is usually faster than the local adaptive method and the global method will give good results for specific weather conditions such as object detection in winter condition. In these cases, acquired images have uniform contrast distribution of foreground and background and little variation in illumination. We are looking for an effective method for object detection in infrared images in challenging conditions such as summer conditions or in an urban environment, where there is a shortage of objects of interest. In these cases, we employed local mean techniques and local variance techniques. The experiment results are presented so that we can better choose which method should be employed or what combination of these previous techniques to employ. In order to minimise computational time of local thresholding technique, we employed a combination of two previous techniques. The algorithm was tested in a Matlab environment and the tested pictures were acquired by RayCam C.A- 1884 and thermoIMAGER 160 cameras.