An Intelligent System Mitigating Vertical Fixed Pattern Noise in CMOS Image Sensors
IEEE Transactions on Instrumentation and Measurement · 2026 · 75 · 2513509-2513509
DOI: 10.1109/tim.2026.3709511 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Transactions on Instrumentation and Measurement, 2026. Verfasst von Wilfred Kisku und 2 weiteren Autor:innen. Die Arbeit nennt 45 Literaturangaben.
Abstract (Originalsprache)
With the design of column-parallel readout for high-resolution and high-speed applications, there is an added vertical fixed pattern noise (VFPN) that is observed. To mitigate this VFPN, a novel edge deployable convolutional neural network (CNN) called FPNrNet has been successfully proposed. This work aims to integrate both the sensor system and a computational unit together to validate a real-time system with infused intelligence near the CMOS image sensor (CIS). It compensates for the variations in column response, thereby improving the overall image quality. The work introduces an architectural solution for the designed and fabricated CIS in the AMS 350-nm OPTO process that targets VFPN reduction, leveraging a CNN integrated with an attention mechanism. The developed model size is designed to be $116\times $ smaller than DnCNN, which is a standard deep neural networks (DNNs) denoising model. The VFPN of the raw image from CIS has an estimated standard deviation of 0.43, which is reduced approximately by $5.5\times $ with FPNrNet. The hls4ml platform is used to deploy FPNrNet onto the FPGA fabric with an empirical study of the model efficacy in terms of precision. The efficacy of FPNrNet has been further validated across diverse edge platforms, including Cortex-M microcontrollers and Raspberry Pi devices, using INT-8 quantized models. The results demonstrate a clear tradeoff between model size, inference latency, and reconstruction performance.
Zitieren
APA
Kisku, W., Kodam, S., & Kaur, A. (2026). An Intelligent System Mitigating Vertical Fixed Pattern Noise in CMOS Image Sensors. IEEE Transactions on Instrumentation and Measurement, 75, 2513509-2513509. https://doi.org/10.1109/tim.2026.3709511
Vancouver
Kisku W, Kodam S, Kaur A. An Intelligent System Mitigating Vertical Fixed Pattern Noise in CMOS Image Sensors. IEEE Transactions on Instrumentation and Measurement. 2026;75:2513509-2513509. doi:10.1109/tim.2026.3709511
BibTeX
@article{Kisku2026,
author = {Kisku, Wilfred and Kodam, Sandeep and Kaur, Amandeep},
title = {An Intelligent System Mitigating Vertical Fixed Pattern Noise in CMOS Image Sensors},
journal = {IEEE Transactions on Instrumentation and Measurement},
year = {2026},
volume = {75},
pages = {2513509--2513509},
doi = {10.1109/tim.2026.3709511}
}
EUVIMED