Morphological Operations in Fundus Image Analysis

It was Luc Vincent [10] who first indicated the possibility of the use of mathematical morphological operations in the analysis of ophthalmic images among some other sample applications. In several aspects [11] he illustrated monitoring microaneurisms via typical morphological operations. In general, Luc Vincent extended the basic morphological approaches to grey-scale images and pointed out the role of appropriate adaptive structuring elements in practical applications [12]. There are several reasons why it is worth to develop the mathematical morphology based approach to investigate ophthalmic image data. The first one concerns the fact that, since the time of the pioneer works of Vincent nobody really implemented the morphological tools for the routine ophtalmic data analysis. Even though there are many attempts to apply the morphology approach to various image data e.g. in mammography and solar studies (see e.g [7,8]). The second one is related to recent studies of fundus images. Traditionally most attention has been focused on the glaucoma and diabetic retinopathy cases and the relevant changes observed for the so called nerve disk and cup areas. Recent studies of fundus images (e.g. [4,5,9]) reveal new important clinical observations outside the above areas. In fundus imaging age-related macula degeneration (AMD) and choroidal neovascularization (CNV) (the major cause of severe irreversible vision loss in most developed countries) are typical cases of sets of objects for pattern/shape recognition problems and are particularly suitable for exploiting morphological tools. Sample fundus images illustrating pathological changes are presented in the Figure 1, below. In the case of neovascularization processes often being due to pharmacology therapy, in nephrology e.g., the monitoring of the shapes and sizes of such changes is of a crucial importance in the current medical treatment. Last but not least is the progress of the observation technology enabling collecting fundus images of higher quality and resolutions. The new techniques of