A Context Sensitive Thresholding Technique for Automatic Image Segmentation
Recently, energy curve of an image is defined for image analysis. The energy curve has similar characteristics as that of histogram but also incorporates the spatial contextual information of the image. In this work we proposed a thresholding technique based on energy curve of the image to find out the optimum number of thresholds for image segmentation. The proposed method applies concavity analysis technique existing in the literature on the energy curve to detect all the potential thresholds. Then a threshold elimination technique based on cluster validity measure is proposed to find out the optimum number of thresholds. To assess the effectiveness of proposed method the results obtained using energy curve of the image are compared with those obtained using histogram of the image. Experimental results on four different images confirmed the effectiveness of the proposed technique.
KeywordsConcavity analysis DB index Energy curve Histogram Segmentation
- 9.Duda, R.O., Hart, P.E., Stork, D.G.: Pattern Classif. Wiley, Singapore (2001)Google Scholar