Image Segmentation for Object Detection

Image segmentation is the most important field of image analysis and its processing. It is used in many scientific fields including medical imaging, object and face recognition, engineering and technology. The major challenge of image segmentation is the over-segmentation caused by noise and incorrect spread of intensity on the object. The main goal of this article is to propose methods improving image segmentation in the case of overlapping objects. Prior to applying different segmentation methods, it is necessary to perform specific pre-processing methods preceding application of further segmentation methods, such as de-noising and adjusting of intensity. The three methods of segmentation that are used in this article include the Watershed Distance Transform, Gradients in Watershed Transform and Region Growing Method. The final result of segmentation by the Region Growing Method can be substantially improved by identification of the intersection of overlapping objects. Individual methods have been applied for microscopic crystal image. All methods were designed in the Matlab environment.