TY - GEN
T1 - Retinal vessel centerline extraction using multiscale matched filter and sparse representation-based classifier
AU - Zhang, Bob
AU - Li, Qin
AU - Zhang, Lei
AU - You, Jia
AU - Karray, Fakhri
PY - 2010/7/21
Y1 - 2010/7/21
N2 - Retina located in the back of the eye contains useful information in the diagnosis of certain diseases. By locating a blood vessel's width, color, reflectivity, tortuosity and abnormal branching, one can deduce the existence of these diseases. In order for this to be achieved, blood vessels first need to be extracted from its background in fundus image. In this paper we propose a new method to extract vessels based on Multiscale Production of Matched Filter (MPMF) and Sparse Representation-based Classifier (SRC). First, we locate vessel centerline candidates using multi-scale Gaussian filtering, scale production, double thresholding and centerline detection. Then, the candidates which are centerline pixels are classified with SRC. Particularly, two dictionary elements of vessel and non-vessel are used in the SRC process. Experimental results on two public databases show that the proposed method is good at distinguishing vessel from non-vessel objects and extracting the centerlines of small vessels.
AB - Retina located in the back of the eye contains useful information in the diagnosis of certain diseases. By locating a blood vessel's width, color, reflectivity, tortuosity and abnormal branching, one can deduce the existence of these diseases. In order for this to be achieved, blood vessels first need to be extracted from its background in fundus image. In this paper we propose a new method to extract vessels based on Multiscale Production of Matched Filter (MPMF) and Sparse Representation-based Classifier (SRC). First, we locate vessel centerline candidates using multi-scale Gaussian filtering, scale production, double thresholding and centerline detection. Then, the candidates which are centerline pixels are classified with SRC. Particularly, two dictionary elements of vessel and non-vessel are used in the SRC process. Experimental results on two public databases show that the proposed method is good at distinguishing vessel from non-vessel objects and extracting the centerlines of small vessels.
KW - Matched Filter
KW - multiscale Gaussian filtering
KW - retinal vessel
KW - Sparse Representation-based Classifier
UR - http://www.scopus.com/inward/record.url?scp=77954635158&partnerID=8YFLogxK
U2 - 10.1007/978-3-642-13923-9_19
DO - 10.1007/978-3-642-13923-9_19
M3 - Conference article published in proceeding or book
SN - 3642139221
SN - 9783642139222
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 181
EP - 190
BT - Medical Biometrics - Second International Conference, ICMB 2010, Proceedings
T2 - 2nd International Conference on Medical Biometrics, ICMB 2010
Y2 - 28 June 2010 through 30 June 2010
ER -