Median filters as a tool to determine dark noise thresholds in high resolution smartphone image sensors for scientific imaging

Igoe, Damien P., Parisi, Alfio V., Amar, Abdurazaq, and Rummenie, Katherine J. (2018) Median filters as a tool to determine dark noise thresholds in high resolution smartphone image sensors for scientific imaging. Review of Scientific Instruments, 89 (1). 015003.

[img] PDF (Published Version) - Published Version
Restricted to Repository staff only

View at Publisher Website:


An evaluation of the use of median filters in the reduction of dark noise in smartphone high resolution image sensors is presented. The Sony Xperia Z1 employed has a maximum image sensor resolution of 20.7 Mpixels, with each pixel having a side length of just over 1 μm. Due to the large number of photosites, this provides an image sensor with very high sensitivity but also makes them prone to noise effects such as hot-pixels. Similar to earlier research with older models of smartphone, no appreciable temperature effects were observed in the overall average pixel values for images taken in ambient temperatures between 5 °C and 25 °C. In this research, hot-pixels are defined as pixels with intensities above a specific threshold. The threshold is determined using the distribution of pixel values of a set of images with uniform statistical properties associated with the application of median-filters of increasing size. An image with uniform statistics was employed as a training set from 124 dark images, and the threshold was determined to be 9 digital numbers (DN). The threshold remained constant for multiple resolutions and did not appreciably change even after a year of extensive field use and exposure to solar ultraviolet radiation. Although the temperature effects' uniformity masked an increase in hot-pixel occurrences, the total number of occurrences represented less than 0.1% of the total image. Hot-pixels were removed by applying a median filter, with an optimum filter size of 7 × 7; similar trends were observed for four additional smartphone image sensors used for validation. Hot-pixels were also reduced by decreasing image resolution. The method outlined in this research provides a methodology to characterise the dark noise behavior of high resolution image sensors for use in scientific investigations, especially as pixel sizes decrease.

Item ID: 53468
Item Type: Article (Research - C1)
ISSN: 1089-7623
Copyright Information: © 2018 Author(s). Published by AIP Publishing.
Date Deposited: 07 Mar 2018 07:40
FoR Codes: 46 INFORMATION AND COMPUTING SCIENCES > 4606 Distributed computing and systems software > 460608 Mobile computing @ 50%
40 ENGINEERING > 4009 Electronics, sensors and digital hardware > 400909 Photonic and electro-optical devices, sensors and systems (excl. communications) @ 25%
46 INFORMATION AND COMPUTING SCIENCES > 4603 Computer vision and multimedia computation > 460306 Image processing @ 25%
More Statistics

Actions (Repository Staff Only)

Item Control Page Item Control Page