MSER feature extraction algorithm is blob detector that identifies region in input greyscale image which does not show variation in size at various values of threshold intensity. The advantage of choosing MSER algorithm is that it can effectively detect regions even in blurred image and dense textured image [41,42]. Ranjan et al., [36] reported that various surface defects in FSW can be identified as white blob in weld image. The identified blobs were classified into voids, grooves, cracks, key-hole and flash defects by analysing the size and orientation of white blob in image. Hence in this work, MSER algorithm has been adopted for feature extraction from weld image. The extremal stable regions are detected by calculating change in area of connected component in image while thresholding image at several levels. The extremal region Ri of an image is detected based on following.