Anirban Mukhopadhyay

Publications

• Image Segmentation for a rippled pattern classification on ion bombarded Silicon surface

  Sandip Sarkar, Anirban Mukhopadhyay, Sofiul Mollick

  RTCT 09, Kolkata; 03/2009

  This paper is to propose a problem specific solution regarding the images generated by Atomic Force Microscope (AFM). When Silicon wafers with rough surface morphology are bombarded at a grazing angle, a ripple pattern is formed on the silicon surface. The nature of the ripple depends on the angle a... [more] This paper is to propose a problem specific solution regarding the images generated by Atomic Force Microscope (AFM). When Silicon wafers with rough surface morphology are bombarded at a grazing angle, a ripple pattern is formed on the silicon surface. The nature of the ripple depends on the angle and the energy of the ion beam. We wanted to study of this ripple pattern requires quantification of the defects by counting the number of branching & truncation. For doing this, the other researchers have previously used different global segmentation algorithms like Otshu’s, Intermeans, Iterated Conditional Modes (ICM) etc [1]. These algorithms are based on the assumption that the image is multimodal and therefore failed to produce desirable result because the images under consideration are unimodal. Here, we propose a technique where, instead of using global image threshold, we use local threshold for every pixel of the image [2]. In this technique, we have developed an algorithm which operates with the following steps: Firstly, we have smoothed the whole image in different scales. Then, from those images, we introduce the polarity image. Polarity is the property of the image which identifies the measure to which the gradient vectors in a certain neighborhood, all points in same direction. Then we take the average of these images and locate the pixels whose values are close to 1. Finally, we perform a morphological thinning operation to get our final image. This process yields much better results than the other global image segmentation algorithms.