Retinal Vessels Segmentation Based on Histogram Equalization Combined with Improving Multi-Scale Line Detection

In the health care field, doctors usually analyse some details presented in medical images to detect some diseases. If the doctor sees some changes related to the shape, colour or size of the retinal blood vessel. The doctor will guess early stage of dangerous sickness like age-related macular degeneration, diabetes, hypertension, arteriosclerosis. It is a good time for the doctor to build a treatment plan as soon as possible. One of the useful retinal blood vessel analysis ways is segmentation. The task should be done before other works are deployed to cure the patient. The paper proposed a method to segment retinal blood vessels by using histogram equalization combined with improving multi-scale line detection. This method uses histogram equalization to improve the quality of the input image, and then improving multi-scale line detection (MLD) technique to enhance the accuracy of vessels detection. Particularly, using the MLD technique to detect large vessels and small ones with adaptive window sizes finally combined. As a result, our method can work effectively to segment more vessels. The method is tested in quality and quantity on publicly available DRIVE datasets with an average accuracy reaching to 0.9515. The result of the proposed method is better than these other methods.