Design and Image Analysis of an APS-Derived CMOS Neuromorphic Model Mimicking Retinal Rod and Cone Pathways

Abstract

Neuromorphic engineering has attempted to mimic the biological retina's signal processing to create hardware-efficient vision systems. In this study, photoreceptor activity and its division into rod (low-light) and cone (bright-light) responses were modeled to simulate the early visual pathway using a CMOS-based technique. This study proposed a CMOS photoreceptor model that combines a 3T Active pixel Sensor (APS) with specialized rod-like low-pass and cone-like high-low-pass temporal filters. The objective was to use hardware-efficient analog circuits to mimic early retinal processing. The generation of rod and cone-dominant images formed from the identical input scene using MATLAB visualization, and pixel-level voltage responses (20 mV-1.8 V) were acquired by Cadence simulation. The rods and pathway maintained temporal brightness and contrast at lower and higher photocurrent levels, respectively. The study showed functional correspondence with biological behavior and established a compact foundation for future retina-inspired neuromorphic vision systems.