Detection of blood vessels in retinal images using two-dimensional matched filters

Abstract

Blood vessels usually have poor local contrast, and the application of existing edge detection algorithms yield results which are not satisfactory. An operator for feature extraction based on the optical and spatial properties of objects to be recognized is introduced. The gray-level profile of the cross section of a blood vessel is approximated by a Gaussian-shaped curve. The concept of matched filter detection of signals is used to detect piecewise linear segments of blood vessels in these images. Twelve different templates that are used to search for vessel segments along all possible directions are constructed. Various issues related to the implementation of these matched filters are discussed. The results are compared to those obtained with other methods.

Detección automática de vasos en retinografías

“…Respecto a la validación con las bases DRIVE y STARE, hemos recurrido a la comparación de los resultados obtenidos por otros algoritmos, que han empleado las mismas bases de datos y que se han validado de forma amplia [8][9][10][11] . El parámetro utilizado para evaluar de forma numérica el resultado del algoritmo es la exactitud, estimada por la razón del número total de puntos correctamente clasificados (suma de verdaderos positivos y verdaderos negativos) por el número de puntos en la imagen dentro del campo de visión.…”

Microaneurysms detection using a novel neighborhood analysis

“…Finally a gaussian-shape matched filter is used to define the MA candidates. This filter is a template matching algorithm that is widely used in the detection of the blood vessels in retinal images [15]. The kernel can be defined by m(x) = − exp (−x 2 /2s 2 ), ∀ |y| ≤ L/2, where L is the length of the vessel segment that has the same orientation and s defines the spread of the vessel intensity profile [15].…”

“…This filter is a template matching algorithm that is widely used in the detection of the blood vessels in retinal images [15]. The kernel can be defined by m(x) = − exp (−x 2 /2s 2 ), ∀ |y| ≤ L/2, where L is the length of the vessel segment that has the same orientation and s defines the spread of the vessel intensity profile [15]. For the vessels detection, the kernel is rotated at all possible vessel orientations and the maximum response from the filter bank is registered.…”

Neural network based detection of hard exudates in retinal images

“…First the candidate OD centres are selected using a combination of mathematical morphology (alternative sequential filters) and regional maxima detection. In a second stage, the vertical blood vessels in the preprocessed image are detected using a two-dimensional matched filter with Gaussian cross-profile in the vertical direction [24]. Then, the Hough transform is used to detect vertical lines in the neighbourhood of each candidate point.…”