The increase in the number of papers in optical image processing (OIP) is indicative of a new era in image acquisition and optical data analysis. This trend is fueling the need and the opportunity for innovative techniques that take advantage of the unique properties of optics [1,2]. In general, OIP is useful in applications in which high parallelism and real-time processing can be effectively realized. OIP is still in its early days, and there are a number of directions in which the field is likely to move in the coming years. Representative examples include optical encryption for data transmission, images, or video streams for information technology security, ranging from biometric authentication over digital image forensics to visual passwords [3]. One prominent suggestion has been to use polarized light and hyperspectral imaging (HSI) as additional degrees of freedom in OIP, and this has been the subject of study of several interesting papers [4–12]. This possibility has attracted much attention; in particular, the recently developed approaches [12–18] in optical polarimetric image processing (OPIP) enable significantly increased sensitivity of target detection and object recognition in multiply scattering or poor-visibility media such as underwater environments or biological tissues. However, only recently have researchers begun to unravel the important role that polarized light and HSI play. Object detection and identification can be significantly improved by combining both polarization and hyperspectral signatures at specific wavelengths [17–19]. In this note, we purposefully give only a brief description of some of the theoretical and experimental developments that have taken place in the area of OPIP.

The development of HSI began in the early 80s, mostly driven by the remote sensing community. Since then, a great deal of effort has gone into this field of research [20,21]. HSI has been used for monitoring biodiversity, studying the ocean, and mapping urban areas, to mention a few examples. HSI represents a label-free optical technology that acquires a stack of images over continuous spectral bands across a wide range of electromagnetic spectra, e.g., from the ultraviolet to short-wave infrared (SWIR) regions. In Fig. 1, we present several examples of hyperspectral signals for everyday objects. These signals are shown in the form of a spectral reflectance, describing the ability of the object to reflect a particular wavelength. For instance, blue light is able to penetrate less than 0.1 mm into the skin, while near infrared light (800 nm) penetrates more than 3 mm. Focusing on vegetation, the minimal reflectance before 700 nm is caused by the chlorophyll pigment, as well as the low reflectance between 400 and 500 nm, a region where the carotenoid pigment (yellow, orange, and red organic pigments) also plays a role. If the banana spectrum is affected by carotenoid and chlorophyll as well, the significant decrease of reflectance after 900 nm is due to water absorption. One may recognize the hemoglobin signature in the skin reflectance (absorption around 550 nm). The hyperspectral signals of wood and road are mostly driven by the structural properties of the medium. Overall, HSI contains interesting information for applications requiring a fine quantification of the properties of materials.

 

Figure 1. Illustrating HSI by the spectral properties of common materials in the visible-near infrared region (from [22–24]). Each material has a unique spectral signature, accessible through HSI, making identification and characterization possible.

Download Full Size | PDF

However, today’s hyperspectral imagers are not sensitive to polarized light effects such as specular reflection at the interface between two media, Rayleigh scattering in the atmosphere, or scattering in turbid water [15]. Focusing on an underwater application, for instance, hyperspectral imagers collect photons resulting from multiple scattering leading to blurred images, which can be improved by considering polarized light, as already studied in previous works using RGB images (Fig. 2) [9].

 

Figure 2. Illustrating polarimetric imaging by enhancing underwater imaging (from [6]). Left: underwater classical RGB image. Right: the same scene viewed through linear polarization measurement and contrast enhancement.

Download Full Size | PDF

It is clear that the combination of HSI and polarimetric imaging can be used to develop classification methods that combine both spectral and spatial information from an object, allowing these methods to identify visually indistinguishable features in a scene within a large region of interest. However, we do not yet know what might be possible for characterizing small scales, especially in disease diagnosis [25] and image-guided surgery, and we want to be open to the broadest range of possibilities to develop the feasibility and usefulness of a polarization-sensitive hyperspectral imager that can be used as a noninvasive instrument [25].

It was a rewarding experience to contribute to the prestigious AOP [1,2]. The submission system was very smooth and reviewer comments were appropriate and useful for improving our manuscript. The process, from initial submission through final acceptance, was really fast.