الفهرس | Only 14 pages are availabe for public view |
Abstract Major characteristic of most past research activities in image coding is their dedication to block-oriented principles, having resulted in a large variety of concepts for still image coding such as the JPEG standard, vector quantization and block truncation coding. This adherence to a fixed block partition strongly neglects the image content. As a consequence, increasing data compression inevitably leads to severe blocking artifacts, ringing effects, and to a rapidly deteriorating image quality. Abandoning fixed block partitions Il1 favor of a content (elated linage representation, a new class of image coding called second generation image coding or segmented image coding has emerged. In this class of coding techniques, the image is segmented into quasi- stationary regions. with uniform texture according to a suitable homogeneity criterion. The obtained description in terms of textures (segments) and contours (boundaries) seems to be more natural in that itcoincides with psychological concepts of vision. In addition, due to the closer adaptation to local image characteristics, one can expect the attainable compression ratios to be significantly higher’ than in block-oriented coding schemes if powerful description and coding tools for contour as well as for texture are at hand. This thesis is devoted to improving and developing new fuzzy logic-based image processing techniques with respect to three main issues.. image segmentation~ edge detection and compression. As far as fuzzy image segmentation is concerned, three clustering-based techniques and two histogram-based techniques are introduced. For fuzzy edge detection, a new minimum entropy-based fuzzy edge detection technique is proposed. With respect to fuzzy image compression, two new systems are introduced. The first one is a fuzzy segmented-image coding system that belongs to the family of the second¬generation image coding. The second one is an adaptive perceptual image transform coding system. |