Scale based image decomposition

An image decomposition method called Scale Based Image Decomposition is proposed m this thesis. Images are decomposed into their different levels of spatial scales in such a way that superposition of the various scales produces a replica of the original image. The method assigns a scale de pendent parameter to edge points in the image. This scale parameter depends mainly on the distance to the closest edge. Edges having the same scale parameter are grouped and considered to represent the same spatial scale. The mam advantages of this process are that it does not require a priori knowledge of the nature of the images to be decomposed and does not introduce any alteration in the size or shape of the image detail. Theoretical development of the Scale Based Image Decomposition technique is presented through a detailed analysis of the behavior of one-dimensional signals processed by the Laplacian of the Gaussian bandpass filter with different bandwidths. The two-dimensional case is also considered. This thesis also contains implementation details together with sample results of decomposed one and two-dimensional functions.