Contrast enhancement using morphological scale space

Contrast enhancement is a necessary pre-processing step in many image processing algorithms. This paper introduces a new contract enhancement algorithm designed specifically for segmentation applications in which an image contains multiple objects of different sizes. The underlying assumption of our algorithm is that an object can be best segmented if it is locally enhanced at a scale that corresponds to the object size. Our method uses a multi-scale image decomposition, obtained with a series of morphological top-hat transformations where the scale of enhancement corresponds to expected object size. In addition, this method is direct where the level of enhancement is controlled using a contrast measure. Finally, our method is adaptive where the enhancement is applied locally, based on local image properties. We demonstrate the effectiveness of our method with experimental results, where we illustrate how our algorithm works and quantitatively measure the improvement in the quality of segmentation, using oil sand images as an example.