Crack image processing using probability based threshold

In this paper a crack image acquisition and processing technology is proposed based on modern digital photography technologies. In order to realize the crack segment from its background without manual interference, a binarization threshold is automatically defined according to the probability distribution between the crack and background. This operation can capture the mostly possible crack pixels against its background. To overcome the limitation of the traditional median filter which is more suitable for salt and pepper noise filtering, a block filter algorithm is proposed to remove the salt and pepper and isolated islands. The experimental results show that block filter algorithm can eliminate both kinds of noises better than the traditional median filter, without polluting the original crack borders.