Retinal Blood Vessel Segmentation using Curvelet Transform and Morphological Reconstruction

Digital images are obtained from the retina and graded by trained professionals. However, a significant shortage of professional observers has prompted computer assisted monitoring. Assessment of blood vessels network plays an important role in a variety of medical disorders. Manifestations of several vascular disorders, such as diabetic retinopathy and hypertensive retinopathy depend on detection of the blood vessels network. The novel vessel segmentation method starts with the contrast adjustment of the green channel image representation to increase the dynamic range of the gray levels, so that the vessels will appear brighter than the background. A multi-scale method for retinal image contrast enhancement based on the Curvelet transform is employed on the contrast adjusted image. The Curvelet transform has better performance in representing edges than wavelets for its anisotropy and directionality, and is therefore well-suited for multi-scale edge enhancement. The Curvelet coefficients in corresponding subbands are modified via a nonlinear function and take the noise into account for more precise reconstruction and better visualization. The morphological operators are used to smoothen the background, allowing vessels, to be seen clearly and to eliminate the non-vessel pixels. The described techniques in this work are applied on images from eye hospital. The proposed algorithm being simple and easy to implement, is best suited for fast processing applications.