I'm working on a project to do segmentation of thermal images. So far, the methods have produced limited segmentation performance (k-means, Otsu's Method, Sobel, Canny, LoG).
I am trying to achieve good segmentation of doors and windows in images like these. What approaches would you suggest to try?
I have tried to convert the images to grayscale using the temperature map, but this has done little to improve segmentation accuracy of these regions of interest.


;), with maybe embedded squares/rectanges embedded within. – Duck Dodgers Jul 26 '19 at 13:22