Contourlet transform and morphological reconstruction based retinal blood vessel segmentation

Abstract

Retinal images are used in ocular fundus operations as well as human recognition. It has significant role in early detection of diabetic retinopathy, hypertensive retinopathy and retinitis pigmentosa by comparing the states of retinal blood vessels. The novel vessel segmentation method presented in this work starts with the contrast adjustment of green channel image representation to increase the dynamic range of the grey levels. A multi-scale method for retinal image contrast enhancement based on Contourlet transform is employed on the contrast adjusted image. The Contourlet transform provides better performance in enhancing the vessel-like segments than the Wavelets and Curvelets with its anisotropy and directionality characteristics, and is well-suited for multi-scale and multi-directional edge enhancement. Contourlet coefficients obtained via the contourlet transform in corresponding sub-bands are modified using a non-linear function. Noise is taken for more precise reconstruction and better visualisation. The proposed algorithm is best suited for fast processing applications.