Design of Probabilistic Patch Based Glaucoma Detection using CDR by Optic-Disc and Optic-Cup Segmentation

Denoising and Texture features (DTF) within eye images are energetically tracked for more precise and efficient classifications of different diseases in glaucoma images which are captured by very sensitive camera lens. Due to improper lens adjustment and noise usually called speckle and Gaussian noises are the major unwanted component pixels present in the image while capturing and these makes difficult for further accurate diagnosis or automatic image interpretation. To address these issues, this paper develops and demonstrates a new approach for removal of speckle and Gaussian noises using Probabilistic Patch Based (PPB) filter, which is extended version of Non Local means algorithm. The PPB filter is depends on the distribution noise model and maximum weighted likelihood estimation and these weights are training iteratively based on the both noisy patches and patches extracted similarities from the previous estimations as part the this work. In the second part, the features are extracted using Lifting Scheme DWT in terms of low and high frequencies into sub-bands and the required low frequency components are used for further analysis. Cup-to-Disc components are segmented using basic level set function and based on the values of cup and disc values, the glaucoma is affected or not is classified. Comparative analysis has made between proposed results and previously available results and it is found that the present results shows 29% is improvement in accuracy and 15% improvement in accurate identification of glaucoma