Segmentation of shallow scratches image using an improved multi-scale line detection approach

Along with developing modern technology, the demands for optical element surface develop towards the characteristics of large scale and high precision. However, it is challenging to evaluate the surface defects since some shallow scratches in optical element surface images are usually characterized by low contrast and blurry outlines. This property makes the machine vision inspection extremely difficult. So, this paper proposes a novel multi-scale line detection method that can efficiently extract shallow scratches. Firstly, to decrease the influence of the surrounding region, a new multi-scale line detector combines all the responses at different scales by setting different weights for each scale. Then, based on the scratches features, we utilize morphological operations to get the full continuum of the scratches area. Experimental results show that our model can ideally extract the contours of shallow scratches that are very close to the optical microscope results observed by specialists.