Improve Image binarization and segmentation

I have a set of videos whose images I wish to binarize such that the background is black and everything else is white. The lighting varys from video to video and the crack grows as video progresses.

I have a function which will crop in on the crack in the sample (ask2 image) but then need to binarize this image as accurately as possible such that the background is black and everything else is white.

Simply thresholding produces innacuracies in different frames (images attached are clearer than most), what is the best method to make this as accurate as possible?

Answer 1

[KALYAN ACHARJYA]

From the attach images, it seems easy, try with global thresholding

Here threshold value is chaning as per image

threshold=graythresh(grayImage);
BW=imbinarize(grayImage,threshold);

Another way, if the dark regions are clipped, then use Histogram Equalization before binarization.