Automatic Segmentation Based on Statistical Parameters for Infrared Images

Image segmentation is an integral part in recognizing pat- terns. Image segmentation techniques aim to partition of images into several parts so that object and background are separated for image un- derstanding and analysis. There are many image segmentation method presented but very few work with infrared images (IR). Fast improving performance and falling cost of IR sensors strengthen IR image process- ing popular. IR images provide more capability to capture images at large distance without light illumination in diverse environment condi- tions, which is not present in visual images. IR image segmentation grows slowly in practical aspect rather than theoretical. Thresholding is sim- ple and widely used method for image segmentation. For this work, IR images are captured using SeekThermal IR sensor. The various statis- tical parameters such as mean, mode, median, standard deviation (SD) etc. are retrieved from input image. Based on these statistical parame- ters a new automatic method for image segmentation is proposed called as StatSDM. The proposed StatSDM method uses combination SD and median for automatic image thresholding. The performance of statSDM is evaluated with standard statistical based image segmentation meth- ods. The results are compared with global Ostu, Max Entropy, Trian- gle and Percentile thresholding techniques show promising performance. This work presents automatic and efficient thresholding method for IR image segmentation.