I am not sure about this but i think this will be the right one ...

MULTISCALE GRADIENT ALGORITHM AND IMAGE SEGMENTATION CAN BE DONE HOPE THIS WILL BE EFFECTIVE IN DETECTION OF CRACKS....

Watershed segmentation may help. or you can design a region growing approach to track the cracks in a binary version or you may also try morphological operations

Dear Mohammed,

the first and primary for crack detection is binary segmentation (thresholding) such as: Kapur, Otsu and other techniques.

if the primary method doesnt work properly, you should use better methods like: fuzzy segmentation, neural nets, etc.

if you would like use the neural networks, I prefer to use the SVM (support vector machine). this technique has high accuracy and less time consuming.

Dear Faisel Mohammed,

Look the link

www.mdpi.com/1424-8220/14/10/19307/pdf

www.niemueller.de/uni/crackdet/crackdet.pd

Regards, Shafagat

Watershed algorithm and Active contours are the best methods for crack detection.

u can find region based connected component then apply morphological filters or other filters.

try : bwareaopen() function in matlab

Dear Dr. Faisel

Since crack region represent the local darkest region, I propose applying the following steps:

1. Apply local brightness stretching (for example Gamma mapping) to enforce the blackness of dark areas.

2. Do binarization using local thresholding (with low threshold values) to construct binary index array, such that the dark pixels (including cracks) are flagged "0" and other regions pixels as "1"s .

3. Apply seed filling filling algorithm on the binary index array to collect the thin black segments (i.e., a set of connected black pixels).

Regards 

You can try Inpainting algorithm or Second Order derivatives. I think they are mainly used to identify and replace lost or corrupted parts of an image data.   

Maybe you develop your own algorithm, making use of mathematical morphology.