Democritus University of Thrace (DUTH)
Γρήγοροι αλγόριθμοι επεξεργασίας και ανάλυσης δυαδικών εικόνων
Abstract
dc:descriptionIn the present PhD thesis, a new binary image representation scheme, which is called Image Block Representation (IBR), is presented. This representation is a region representation method. In the Image Block Representation, the whole binary image is decomposed into a set of rectangular areas with object level, which are called blocks and have their edges parallel to the image axes. Each pixel of the image with object value belongs to one and only one block. The Image Block Representation is an information lossless process and therefore, from an information theory perspective, it is equivalent to the 2-D array image representation. In comparison with other region based representations schemes and specifically with the run length coding and the quadtree representation, the IBR is superior, since the number of the blocks is significantly smaller than the number of runs and the number of quadtree nodes with object level, respectively. As in every other region based method, the most important characteristic of the image block representation is that a perception of image parts greater than a pixel, is provided to the machine and therefore, all the operations on the pixels belonging to a block may be substituted by a simple operation on the block. Taking this feature into account, the implementation of new algorithms for binary image processing and analysis tasks, leads to substantial reduction of the required computational complexity. In this doctoral thesis, a number of basic image processing and analysis algorithms, which are implemented on block represented images are presented. These algorithms are characterized by low time complexity and are suitable for very fast processing rates, due to the substitution of image pixels by blocks. Specifically, these algorithms are: The fast implementation of the image shift, image scale, image rotation operations. The connected component labeling, operation that permits the detection of objects and the determination of the connectivity among image areas. The determination of the minimum and of the maximum distance from a point to an object, edge extraction, perimeter measurement and area measurement. The fast implementation of logic operations among blocks and images. The implementation of the Set Difference operator which is used as an auxiliary operator for the execution of the OR, XOR and NOT operations in block represented images is also presented. The real-time computation of statistical moments. Moments constitute a well-known tool for image analysis and have extensively used for pattern recognition tasks. However they suffer from a high computational cost. By exploiting the rectangular structure of the blocks, the moments computation with a rate greater than 50 fps (i.e. real-time) using a low cost serial computer, is achieved. The IBR process and the real-time moments computation on block represented binary images, using the parallel architecture Scan Line Array Processor (SLAP). SLAP is a linear array of N processors, operates in a SIMD (Single Instruction Multiple Data) mode and has been used for image processing tasks. Two fast algorithms for the moments computation on gray images. The first method provides the exact moment values and operates much faster than the conventional moment calculation and faster than other fast moment calculation technique. The second method provides approximated moment values, with small error of 2-3% and operates in real-time. The thinning and skeletonization of objects. The object’s skeleton provide useful information for its shape, which is also considered as independent from noise and from the scale of the object. Two thinning algorithms are presented in the thesis. Finally, some applications of the above algorithms are also presented.
Degree
thesis:*- Grantor dc:publisher
- Democritus University of Thrace (DUTH)
- Year dc:date
- 2007
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Spiliotis, Iraklis
- Σπηλιώτης, Ηρακλής
Subjects
dc:subject × 16- Επεξεργασία εικόνας
- Ανάλυση εικόνων
- Γρήγοροι αλγόριθμοι
- Δυαδική εικόνα
- Υπολογισμός ροπών
- Αναπαράσταση με ομάδες
- Image processing
- Image analysis
- Fast algorithms
- Binary image
- Image block representation
- Moment computation
- Επιστήμες Μηχανικού και Τεχνολογία
- Επιστήμη Ηλεκτρολόγου Μηχανικού, Ηλεκτρονικού Μηχανικού, Μηχανικού Η/Υ
- Engineering and Technology
- Electrical Engineering, Electronic Engineering, Information Engineering
Rights
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/20923