Iterative elimination algorithm for thermal image processing
Segmentation is employed in everyday image processing, in order to remove unwanted objects present in the image. There are scenarios where segmentation alone does not do the intended job automatically. In such cases, subjective means are required to eliminate the remnants which are time consuming especially when multiple images are involved. It is also not feasible when real-time applications are involved. This is even compounded when thermal imaging is involved as both foreground and background objects can have similar thermal distribution, thus making it impossible for straight segmentation to distinguish between the two. In this study, a real-time Iterative Elimination Algorithm (IEA) was developed and it was shown that false foreground was removed in thermal images where segmentation failed to do so. The algorithm was tested on thermal images that were segmented using the inter-variance thresholding. The thermal images contained human subjects as foreground with some background objects having similar thermal distribution as the subject. Informed consent was obtained from the subject that voluntarily took part in the study. The IEA was only tested on thermal images and failed when false background object was connected to the foreground after segmentation.