Multiscale Morphological Techniques for Image Processing

In this paper we have shown how a common multiscale morphological paradigm can be used to solve various image processing problems. We have implemented the schemes by using morphological tower which is based on multiscale operations satisfying causality and edge localization properties. The efficacy of the technique is illustrated for noise smoothing, contrast enhancement and multimodal image fusion problems. Introduction Image processing involves various types of application specific operations on acquired images. In general, an image after its formation may not be suitable for subsequent processing or analysis due to various reasons like noise contamination, poor contrast and representation style [2]. Various types of processing are carried out to reduce these problems. Due to difference in response, a sensor fails to visualize all the objects of the same scene with equal clarity. The image formed using such a sensor therefore fails to depict all salient and subtle features. The images of the same scene recorded by different type of sensors may be fused to obtain the image having better visual clarity of all such objects. Another important processing is segmentation where the image is decomposed into a set of meaningful regions (e.g. objects and background). Mathematical Morphology is a potential tool for dealing with shapes. Morphological operations, namely erosion, dilation, opening, closing etc. are used for extracting, modifying, manipulating the features present in the image based on their shapes. Morphological operations are by and large set theoretic operations defined between two sets: the object and the structuring element (SE) [11]. The shape and the size of SE play a crucial role in such type of processing and are therefore chosen according to the need and purpose of the associated application. In general, an image comprises of objects or features of different sizes or scales. On the other hand, image processing techniques often make use of some pre-specified parameter values. Indeed these parameter values, when chosen globally, irrespective of scales of the objects or features, may not do justice to them equally. Thus the need for processing the image based on the size or scale of objects or features present in the image has initiated several scale-space techniques. When the concept of scale is introduced into the mathematical morphology through the structuring element, the result is multi-scale morphology [11,4]. In this work our aim is to show the application of multi-scale morphological techniques to various image processing problems.