Underwater imaging has always been a challenge due to limitations imposed by scattering and absorption nature of the underwater environment. The light would be highly degraded after reflection and propagation in the water medium. Being an advanced imaging technique, Single-pixel Imaging (SPI) is applicable to acquire object spatial information in low light, severe backscattering, and high absorption conditions. Combination of Compressive Sensing (CS) and SPI can overcome the limitation of SPI algorithms such as long data-acquisition time, low reconstruction efficiency and poor reconstruction quality. In the current research, an underwater SPI system based on CS is established to reconstruct our two-dimensional (2D) transparent object. We have systematically investigated the influence of water turbid degree, measurement pattern types and number of measurements on image reconstruction performance. The proposed system is capable to reconstruct the object even when the turbidity reaches up to 80 Nephelometric Turbidity Unit (NTU), where the conventional imaging systems are unusable. Proposed reconstruction method in our research can save more than 70% data acquisition time, compared to SPI algorithm. Our experimental setup has been compared to a conventional imaging system and an underwater ghost imaging system to show its efficiency in obtaining accurate results from turbid water conditions. Furthermore, various algorithm comparison and imaging enhancement studies demonstrates that our algorithm is superior in bringing highly convex optimization at a faster rate with a smaller number of measurements. This work creates new insight into the SPI application and generates a guideline for researchers to improve their applications.

Introduction

Imaging in an underwater environment is difficult particularly in turbid media due to backscattering and absorption effects. Various researches have been done aiming to capture the images of objects in underwater environments [1], [2] and underwater imaging enhancement [3], [4]. Due to the presence of faster, more dynamic computers and strong fundamental spatial light mapping mechanisms such as Spatial Light Modulator (SLM) and Digital Micromirror Device (DMD) developed in the last decade comprehensive investigations have been done in the field of imaging which employs only a single pixel detector noted as Single Pixel Imaging (SPI) [5]–[۷]. Compared to the conventional method, SPI makes it possible for the imaging system to use a low-cost single-pixel photodetector rather than an expensive multi-pixel Complementary Metal Oxide Semiconductors (CMOS) or Charge-Coupled Devices (CCD). It is convenient and efficient to manufacture single-pixel detectors with a large active area, which makes SPI techniques more appropriate for imaging, especially in low illumination conditions (underwater and foggy medium, etc.) than traditional imaging techniques. The limited response of detection medium to a certain bandwidth has been a flaw of conventional imaging systems. The same drawback of the conventional system has become the major advantage of SPI because it can make measurement relatively in a much wider range of the electromagnetic spectrum.