Curvilinear Object Detection with Fuzzy Mathematical Morphology for Grayscale and Color Medical Imagery
[eng] Fuzzy mathematical morphology is a set of tools to process grayscale images. It is based on two operators, the dilation and the erosion, that respectively enlarge and shrink objects. We extend these operators to deal with multivariate images by defining the soft color dilation and the soft col...
| Autor: | |
|---|---|
| Tipo de recurso: | tesis doctoral |
| Estado: | Versión publicada |
| Fecha de publicación: | 2018 |
| País: | España |
| Institución: | CBUC, CESCA |
| Repositorio: | TDR. Tesis Doctorales en Red |
| OAI Identifier: | oai:www.tdx.cat:10803/665935 |
| Acceso en línea: | http://hdl.handle.net/10803/665935 |
| Access Level: | acceso abierto |
| Palabra clave: | Image Processing Curvilinear Object Detection Vessel Segmentation Fuzzy Mathematical Morphology Soft Color Morphology Lògica borrosa i fusió de la informació 004 51 |
| Sumario: | [eng] Fuzzy mathematical morphology is a set of tools to process grayscale images. It is based on two operators, the dilation and the erosion, that respectively enlarge and shrink objects. We extend these operators to deal with multivariate images by defining the soft color dilation and the soft color erosion. They are designed for generic multivariate color spaces, but also to process natural images consistently with regard to the notions of enlarging and shrinking objects. Besides being able to preserve colors, other theoretical properties are transferred from the fuzzy mathematical morphology. The soft color dilation and erosion can also be combined, in the same way as the fuzzy erosion and dilation, to provide operators with a complex behaviour. Several of such combinations have been designed for a variety of tasks, and can now be transferred to color images: noise filtering, contrast enhancing, object segmentation and shape recognition, among others. In this thesis, we also propose a definition of curvilinear objects to unify the literature: several image processing problems consider the task of segmenting tubular-shaped objects clearly different to their surrounding background. In particular, we study such problems to extract their common denominator. This state of the art is synthesized by categorizing both the approaches to segment curvilinear objects and the features they consider of interest. Besides, we design algorithms based on morphological operators to segment curvilinear objects. We use fuzzy mathematical morphology to segment vessels in eye-fundus photographs and soft color morphology to detect hair in dermoscopic images. Both morphologies consider different implementations of erosion and dilation. However, the dilation and erosion of each morphology can be combined similarly. Both methods achieve high performance compared to other published works. This has several implications: first, it indicates that the soft color morphology is a comprehensible extension of the fuzzy mathematical morphology; second, it is a promising example of the potential of the soft color morphology; and third, it implies that the common denominator of both tasks is extensive enough to face them with similar tools: curvilinear object detectors |
|---|