Democritus University of Thrace (DUTH)
Ανάπτυξη τεχνικών μείωσης των χρωμάτων σε εικόνες
Abstract
dc:descriptionThis PhD presents new methods for the reduction of colors in gray and color images, as well as new methods for the estimation of the number of image's dominant colors. Most of the proposed methods are based on powerful self-organized neural network classifiers. More analytically, the following techniques are developed: For gray-scale images, three new methods are presented for the estimation of the number of image dominant colors, which are based on histogram information. In addition, extensions of these techniques are proposed for color images that utilize the Principal Components Analysis (PCA) to construct histograms that allow the estimation of the number of principal colors. For color reduction of gray-scale images, a new method is proposed where the gray-level value of each pixel is related to local spatial features extracted from its neighboring region. Doing this, the one-dimensional histogram-clustering approach is converted to a multidimensional feature-clustering technique. The proposed technique is based on a powerful classifier, which consists of a PCA neural network and a Kohonen Self-Organized Feature Map neural network (Kohonen SOFM). To improve the subsampling procedure and to speed up the entire algorithm, a fractal scanning sub-sampling technique is applied which is based on the Hilbert's space filling curve. A new color reduction method for color images, the Adaptive Color Reduction method (ACR), is presented which is based on the PCA-Kohonen SOFM classifier. The proposed technique consecutively split and merges the produced color classes in such a way that the initial image to be adaptively decomposed into a number of color planes. Applying this tree driven splitting and merging procedure, optimal quantization results are achieved and the number of final classes are automatically estimated. A new method for Page Layout Analysis of color documents is presented that uses the ACR color reduction technique. Specifically, using the ACR technique, the initial color document is decomposed into a proper number of color planes. In each color plane the position of text is obtained and finally, using a merging procedure, the document's text areas are identified.
Degree
thesis:*- Grantor dc:publisher
- Democritus University of Thrace (DUTH)
- Year dc:date
- 2004
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Atsalakis, Antonios
- Ατσαλάκης, Αντώνιος
Subjects
dc:subject × 20- Μείωση χρωμάτων
- Κβαντισμός χρωμάτων
- Αυτοοργανούμενα νευρωνικά δίκτυα
- Κύριες χρωματικές συγκεντρώσεις
- Αυτοοργανούμενος πίνακας απεικόνισης χαρακτηριστικών του Kohonen
- Ανάλυση κυρίων συνιιστωσών
- Αναπτυσσόμενο νευρωνικό αέριο
- Ανάλυση δομής έγχρωμων εγγράφων
- Color reduction
- Color quantization
- Self organized neural networks
- Image's dominant colors
- Kohonen self-organized-feature-map
- Principal component analysis
- Growing neural gas
- Page layout analysis of color documents
- 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/17441