Adaptive Bilateral Filterative Morlet Histogram Thresholding Segmentation Based Multi-Layer Classification for Early Glaucoma Disease Detection
Accurate and early disease detection plays a vital role due to the development of image processing in biomedical and healthcare communities. Retinal imaging is a new eye test used to identify many diseases. Retinal images play an essential part in diagnosing many ocular diseases like diabetic retinopathy, Glaucoma and Stargardt disease. But the conventional technique faces major challenges to achieve disease diagnoses with lesser complexity. An Adaptive bilateral filterative Morlet Histogram Thresholding Segmentation based Multi-Layer Log-Linear Classification (ABFMHTS-MLLC) technique is introduced for early detection of disease with higher accuracy and lesser time consumption. The ABFMHTS-MLLC technique comprises more than three layers for detecting the Glaucoma disease at early stage. An input image is given to input layer. After that, an input image is sent to the hidden layer 1. In that layer, an adaptive bilateral filter is applied to remove unwanted noise from the input image for early detection of disease. Then, preprocessed image is sent to the hidden layer 2 where Morlet Transformed Feature Extraction is carried out to extract the relevant features such as color, texture and intensity of retinal fundus image. Then, extracted features are sent to the hidden layer 3 for performing the segmentation process. Fuzzy Histogram Thresholding Segmentation method is used to divide the image into number of parts by analyzing the peaks and valleys. Finally, the segmented image is sent to the output layer where Log-Linear analysis is carried out to evaluate the relationship between testing image and training image for classifying the retinal fundus images into normal or glaucomatous. Experimental evaluation is carried out using retinal fundus image database on factors such as peak signal-to-noise ratio, disease detection accuracy, error rate and disease detection time with respect to number of retina fundus image and image size. The ABFMHTS-MLLC technique provides the promising results in terms of achieving higher accuracy and lesser error as well as minimum time consumption than the existing methods.
Keywords: Glaucoma Disease detraction, retina fundus image, preprocessing, feature extraction, Morlet Transformed Feature Extraction, Fuzzy Histogram Thresholding Segmentation, classification, Log-Linear analysis