Image Preprocessing Assessment Detecting Low Contrast Regions under Non-homogeneous Light Conditions

This paper focuses on evaluating the pre-processing impact in detecting low contrast regions on irregular surfaces with non-homogeneous lighting. Non homogeneous lighting represents an obstacle to the correct segmentation and subsequent classification of relevant image regions. For example in grayscale images, intensity variations are detected on the same region. Therefore lower contrast regions require an adequate sensitivity level at the segmentation stage. Segmentation, description and classification techniques will be applied over a set of images without pre-processing and over the same set of images with pre-processing, in order to achieve the assessment. The images used in this paper were obtained from a visual inspection prototype for flaw detection on dentures. The outcome shows that an appropriate image pre-processing is required to improve the detection process performance for the given circumstances.