Nonlinear shape-based image analysis and coding

Shape is a crucial descriptive feature characterizing objects of interest in images. This research addresses nonlinear shape-based image analysis and coding. Efficient shape-based image analysis and coding rely on a prior robust shape representation and object-based segmentation. Consequently, the problem of nonlinear shape-based image analysis and coding addressed in this dissertation leads to main sub-problems, namely shape representation, object-based segmentation, and coding.; A novel morphological reversible contour representation of discrete binary images is proposed. A binary image is represented by a set of non-overlapping multi-level contours and a residual image. In this proposed representation, the total number of pixels representing an image is far less than the total number of pixels obtained by the seed-based morphological contour-skeleton lossless representation. The proposed contour representation is simple, unique, and general without restrictions on the binary image to be represented. Moreover, it requires less number of operations to compute the proposed representation compared with other lossless morphological representation methods. The resulting multi-contour image component is also robust to noise. An efficient differential chain contour coding scheme is employed to further compress the represented image. The proposed method yields very low bit rates compared to the existing morphological techniques. To exactly reconstruct an original image, a new automatic filling procedure, which properly fills a proper multi-contour image according to its topological structure without need of seed points, is proposed. The morphological unique contour representation and its lossless reconstruction techniques have been tested on images with varying size and complexity.; A new object-based segmentation algorithm is proposed in which text regions are the object of interest to be segmented. Compared with existing methods for text extraction from images, the proposed morphological technique is insensitive to noise, skew and text orientation. It is also free from artifacts that are usually introduced by both fixed/optimal global thresholding and fixed-size block-based local thresholding.; Novel object/region-based image representation and coding techniques are proposed. A gray-level image is represented by an edge image, binary mask, and residual (error) image. The edge image is extracted using a nonlinear edge detection technique. The binary mask identifies the different regions/objects in the original image. The extracted edge and binary mask images are insensitive to noise. The residual image is the difference between the original image and an edge-interpolated image obtained by interpolating the edge image using a novel 2-stage nonlinear interpolation algorithm. The proposed edge-interpolation algorithm outperforms the existing morphological interpolation technique. The interpolated image preserves the perceptually significant features in the original image. The binary mask, edge and residual images are coded using the proposed reversible contour coding technique and predictive/entropy coding-based and Set Partitioning in Hierarchical Trees-based coding techniques, respectively. For lossless coding, an LZW-based method is presented. The proposed representation and coding methods provide flexible object/region-based decoding and accessibility. Arbitrary-shaped object (s)/region(s) of interest in images can be coded at low bitrates more efficiently using the proposed coder than using the SPIHT coder.