Author Type

Graduate Student

Date of Award

Spring 5-8-2026

Document Type

Thesis

Publication Status

Version of Record

Submission Date

May 2026

Department

Computer and Electrical Engineering and Computer Science

College Granting Degree

College of Engineering and Computer Science

Department Granting Degree

Electrical Engineering and Computer Science

Degree Name

Master of Science (MS)

Thesis/Dissertation Advisor [Chair]

Hari Kalva

Abstract

I present a thorough noise characterization of the Canon MS-500, a Single-Photon Avalanche Diode (SPAD) camera system, tested under both lit and dark conditions. The camera outputs 10-bit digital number (DN) values produced by an internal processing pipeline whose design is not publicly documented. All analyses in this thesis therefore describe the camera’s DN output — the signal that any downstream detection, tracking, or classification system will actually receive — rather than the photon-counting statistics of the underlying SPAD array. All computations were performed on the native 10-bit data. Where a measured quantity has a known photon-counting analog, the relationship is noted while the measurement itself is reported in DN-domain terms.

Under illumination, I measured a signal-to-noise ratio (SNR) of 11.3 dB, a total noise RMS of 133.6 DN, and a fixed-pattern noise (FPN) RMS of 27.3 DN. The scene-wide mean variance-to-mean ratio (VMR) of 9.4 in DN indicates excess output variance relative to a Poisson reference (VMR = 1), consistent with spatially correlated output between neighboring pixels. The 3 × 3 inter-pixel correlation matrix shows Pearson coefficients up to 0.535 in the horizontal direction. Only 27 hot pixels (0.001%) were found under illumination. The temporal power spectral density (PSD) follows a 1/f slope of −1.39, pointing to dominant low-frequency flicker noise from trap states. The photo-response non-uniformity (PRNU) of 24.2% reflects the illumination pattern rather than pixel-level gain differences. Principal component analysis (PCA) and independent component analysis (ICA) separate scene content from noise into clear structures. The first principal component captures only 5.6% of variance, confirming that the noise field is spread across the sensor.

In the dark, I measured a per-frame temporal noise RMS of 29.0 DN, a dark signal non-uniformity (DSNU) of 30.0 DN RMS, and a mean dark signal rate of 949.85 DN/s relative to the per-frame mode. A total of 258,903 pixels exceed the 3-sigma dark threshold. Cross-dataset validation of all 77 pixels exceeding the 50-sigma threshold against the flat-field and illuminated scene datasets found no pixels persistently bright across all three conditions at that threshold; however, patch-level comparison of the 10 brightest dark pixels shows that 6 of those 10 remain above baseline in the scene dataset, indicating persistent trap-state behavior in the highest-emission tier, while the remaining 4 are RTS switchers whose elevated output was confined to the dark acquisition window. The lag-1 autocorrelation of 0.72 and frame-pair Pearson correlation of 0.80 confirm strong inter-frame persistence in the DN output, consistent with trap-assisted burst behavior. The dark noise distribution has a kurtosis of 45.4 and skewness of 5.3 — far from Gaussian — and exhibits a 1/f flicker character consistent with trap-state dynamics. Together, these results form a complete DN-domain noise budget for this SPAD camera system and provide a methodology that can be adapted for future SPAD characterization work.

Download

Share

COinS