A NEW ERA FOR RETINAL BLOOD VESSEL SEGMENTATION USING SUPERVISED & UNSUPERVISED LEARNING METHOD
Notice: Undefined variable: link_article in /home/ijtc/public_html/plugins/content/bt_socialshare/bt_socialshare.php on line 818
Notice: Undefined property: stdClass::$image in /home/ijtc/public_html/plugins/content/bt_socialshare/bt_socialshare.php on line 818
Notice: Undefined variable: title in /home/ijtc/public_html/plugins/content/bt_socialshare/bt_socialshare.php on line 818
Notice: Undefined offset: 1 in /home/ijtc/public_html/plugins/content/bt_socialshare/bt_socialshare.php on line 820
Notice: Undefined variable: pdFileDate in /home/ijtc/public_html/components/com_phocadownload/views/file/tmpl/default.php on line 277
A B S T R A C T
Segmentation of blood vessels in retinal images used for the early diagnosis of retinal diseases such as hypertension, diabetes and glaucoma. The high resolution, variability in vessel width, brightness and low contrast make vessel segmentation as difficult task. There exist several methods for segmenting blood vessels from retinal images. However, most of these methods fail to segment high resolution (large in size) images, very few methods provide solution for such a high resolution images but it require lengthy elapsed time and the accuracy of these methods is not completely satisfactory. The research work presents a retinal vessel segmentation algorithm which uses a text on dictionary to classify vessel/non-vessel pixels. However, in contrast to previous work where filter parameters are learnt from manually labeled image pixels our filter parameters are derived from a smaller set of image features that we call key points. In the paper work performed to implement the image preprocessing steps such as cropping, color space transformations, channel extraction, color enhancement, gabor filter. After that, preprocessed images will be used to form feature vector and then apply k-means clustering to form segments for blood vessels segmentation. Finally, to compared the existing technique and the proposed technique using the parameters given by accuracy, sensitivity, specificity and thresholding under Receiver operating characteristic (ROC) curve for true positive rate and false positive rate.
Gabor filter, retinal images, segmentations, k-mean clustering, drive dataset, stare dataset.