Eliminating Vertical Fixed Pattern Noise in CMOS-Based Endoscopic Images using Modified Dark Frame Subtraction

Abstract

Complementary Metal Oxide Semiconductor (CMOS) image sensors are widely used in medical endoscopy because of their small size, low power consumption, and high sensitivity. However, they can suffer from vertical fixed pattern noise (VFPN), which appears as a series of vertical stripes in images and can obscure fine details, reducing image quality. Dark frame subtraction is a common technique used to eliminate VFPN, but it may result in poor image quality if the sensor parameters, such as gain and black level clamp (BLC), are not considered. This paper proposes a modified dark frame subtraction method that includes a simple preprocessing step to minimize VFPN by taking into account the impact of sensor gain and BLC. To evaluate the proposed method, it was tested in a simulation and real-time set up and its performance is compared to that of the actual dark frame subtraction. The results show that the modified method significantly improves noise reduction and enhances image quality, making it more suitable for endoscopic applications. By considering the effects of sensor parameters on VFPN, the proposed method offers a more effective solution for eliminating VFPN in CMOS image sensors used in medical endoscopy. This method can help clinicians more accurately diagnose medical conditions by providing clearer images with reduced noise and improved quality.