RETINAL IMAGE SEGMENTATION USING SHEARLET TRANSFORM FEATURES

Retinal vessel segmentation is considered as a major concern for the detection of several retinal disease, at which the manual segmentation is a tedious one. Thus, automatic segmentation of those vessels could aid better diagnosis of retinal disease. As there were several researches carried out on the automatic mechanism for retinal disease, there were some limitations like lower accuracy, reduced prediction rate and so on. An effective CNN based classification technique is employed for the classification and detection of diabetic disease in retinal image. The input dataset is pre-processed by the extraction of green channel with enhancement of image using histogram-based approach followed by filtering to remove noise. The vessel extraction is made using thresholding approach. The pre-processed image is then segmented using Shearlet segmentation algorithm. From the segmented image, the features are extracted by GLCM technique followed by PCA for feature selection algorithm. The selected features are classified and the abnormality is detected. At last, the performance analysis is carried out and is compared with existing techniques to prove the effectiveness of proposed mechanism. The three publicly available DRIVE, STARE and CHASE_DB1 databases are used for performance evaluation.

Built With

• matlab