Comparative signal to noise ratio as a determinant to select smartphone image sensor colour channels for analysis in the UVB

• Published in: Sensors and actuators. A. Physical. Vol. 272; pp. 125 - 133
• Main Authors: Igoe, D. P, Parisi, A.V., Downs, N.J., Amar, A., Turner, J.
• Format: Journal Article
• Language: English
• Published: Lausanne Elsevier B.V 01.04.2018; Elsevier BV
• Subjects: Adaptive algorithms; Adaptive filters; Adaptive threshold; Algorithms; Channels; CMOS; Color; Complementary metal oxide semiconductor; Denoising algorithm; Image integrity; Image processing systems; Irradiance; Light; Mathematical analysis; Metal oxide semiconductors; Scientific visualization; Signal to noise ratio; Smartphones; Studies; Ultraviolet radiation; Ultraviolet-B; Signal-to-noise ratio; Ultraviolet-B; Adaptive threshold; Complementary metal oxide semiconductor; Denoising algorithm; Image integrity
• ISSN: 0924-4247; 1873-3069
• DOI: 10.1016/j.sna.2018.01.057

•Signal to noise ratio quantified for solar ultraviolet B smartphone imaging.•Comparative signal to noise ratio used to compare colour channel integrity.•Signal integrity gauged for chromatic ratios usually used for cloud identification.•Colour channel signal integrity analysed for increasing solar irradiance.•Python data analysis libraries effective in visualising signal integrity. The signal to noise ratio (SNR) is an important consideration for any scientific image sensor application, particularly the relatively low light involved with observations of the solar disc at a discrete ultraviolet-B (UVB) wavelength using an unmodified smartphone image sensor. In particular, the SNR of each of the primary image sensor colour channels (red, green and blue) is a critical step in determining which colour channel signal to analyse for any characterisation research. In each image, the solar disc appears as a very small pale-magenta dot. In this paper, the SNR of each colour channel response for solar UVB, alongside their chromatic transforms were analysed for a stacked, mosaic filtered, backside illuminated complementary metal oxide semiconductor (CMOS) image sensor. Using data visualisation techniques, it has become clear that specific colour channels, in this case – the red channel, provide the strongest SNR for use in characterisation and other analytical research. The effects of a straightforward adaptive threshold and de-noising algorithm (median filter) on each colour channel’s SNR are also analysed. The variation of the colour channels’ SNR with external factors, including irradiance, is modelled. The effects of the prevalence of noise features, such as hot pixels and dark noise, are also observed. It has been found that before the median filter is applied, most of the signal, particularly for the green colour channel, is from these noise features in some image sensors – representing a ‘false positive’ in these low-light conditions. A chrominance model using a weighted proportion of the red and blue colour channels that provides the best SNR when sensing in the UVB waveband for the sensor has been developed and evaluated.