Low-Cost Spatio-Temporal Algorithm for Defective Pixel Identification and Correction

A novel method and hardware architecture is proposed for identifying and correcting defective pixels in CCD or CMOS image sensors. Pixel defects may arise during manufacturing of solid-state silicon sensors, or during use due to high-energy particle impacts. To improve the quality of raw images captured by consumer grade image sensors, most digital cameras, cell phone, surveillance and machine vision cameras use an Image Signal Processing (ISP) pipeline. ISP and digital back end processing modules compete for limited external memory bandwidth and access to frame buffer memory. The proposed architecture supports static identification, can dynamically identify pixel defects without accessing a frame buffer, and can detect stuck and flicker pixel defects of monochrome or color sensors.