Noise-robust low-contrast retinal recognition using compression-based joint wavelet transform correlator

• Noisy low-contrast retinal recognition using CBJWTC is studied via Matlab simulation.
• Retinal target with different noise levels and reference images are JPEG-compressed.
• Low frequency contents of JPS are enhanced by appropriately dilated wavelet filter.
• Simulation results show that the recognition performance is robust to noise.
• The recognition is independent upon compression qualities of the retinal images.

A new method is proposed for recognizing noise corrupted low-contrast retinal images that employs joint wavelet transform correlator with compressed reference and target. Noise robustness is achieved by correlating wavelet-transformed retinal target and reference images. Simulation results show that besides being robust to noise, its recognition performance can become independent upon compression qualities when low spatial-frequency components of joint power spectrum are enhanced by appropriately dilated wavelet filter.