Hypertension Classification for Fundus Image Based on Improving Clahe Morphology in Wavelet Transform and ResUNet

The wavelet transform gives a multilevel for the input stage of feature extraction. The disadvantage of the wavelet domain is the threshold to become the key for the inversion period. The lack of information in medical images is a vital issue for feature maps. Therefore, the input shape of the digital signal needs to improve intelligibility with the background. Machine learning or deep learning cannot easily solve this problem with the basic backbone for feature extraction. In this research, the proposed method for hypertension classification based on the morphology concept in wavelet transform presents another way to parallelize feature maps. The proposed method includes: the pre-processing for parameters by CLAHE morphology, the decomposition for multilevel of parameter input, ResUNet for extraction the feature maps, the inversion to synthesis the feature maps for hypertension classification in wavelet domain. The final results were compared with the recent methods by Accuracy, Sensitivity and Specificity in STARE dataset. These values of comparison are 95.75%, 93.88% and 94.16%, respectively.