Sub-nyquist sampling strategy for compressive single-pixel imaging

Chin, Vui Nam (2022) Sub-nyquist sampling strategy for compressive single-pixel imaging. Final Year Project, UTAR.

Preview

PDF Download (5Mb) | Preview

Abstract

Singe-Pixel Imaging (SPI) captures image using only a single-pixel detector instead of pixelated sensors sensor. It has been a cost-effective alternative to conventional camera especially in non-visible wavelengths and low light conditions. Generally, SPI works with compressive sensing (CS) to achieve sub-Nyquist sampling where efficiency can be increased significantly. Nevertheless, image quality remains a primary concern of SPI. Therefore, this project is aimed to propose an adaptive sampling scheme to improve image quality and efficiency in compressive SPI. Coarse-to-fine (CTF) is able to improve the image quality by sampling the scene progressively from low to high resolution. In this project, in-depth investigation was performed on CTF, particularly to analyse the performance of various step size strategies and progressive reconstruction. Additionally, 1-bit CS was also explored for its potential in performance improvement. Results show that the performance of decreasing step size is better than fixed step size in CTF. Furthermore, outline enhancement is proposed by integrating CTF sampling and 1-bit CS. Based on the results, outline enhancement is able to improve the image quality in SPI especially when the modified sampling patterns is 5 % of total sampling patterns. In conclusion, this project has proposed two sampling methods: decreasing step size as adaptive step size strategy for CTF sampling and outline enhancement scheme to improve the overall image quality in compressive SPI.

Actions (login required)