The development of ultra-compact handheld hyperspectral imagers has been impeded by the scarcity of small widefield tunable wavelength filters. The widefield modality is preferred for handheld imaging applications in which image registration can be performed to counter scene shift caused by irregular user motions that would thwart scanning approaches. Conventional widefield tunable filters like the liquid crystal tunable filter and acousto-optic tunable filter achieve narrow passbands across a wide spectral range by utilizing large interaction lengths, thereby increasing the thickness of the device along the optical path. In addition, these technologies rely on rather bulky external control circuitry and, in the case of acousto-optic filters, high power requirements. In the work presented here, we introduce a novel widefield tunable filter for visible and near infrared imaging based on surface plasmon coupling that can be miniaturized without sacrificing performance. The surface plasmon coupled tunable filter (SPCTF) provides diffraction limited spatial resolution with a <10nm nominal passband and a spurious free spectral range of more than 300nm. Acting on the π-polarized component, the device is limited to transmitting 50 percent of unpolarized incident light. This is higher than the throughput of comparable Lyot-based liquid crystal tunable filters that employ a series of linear polarizers. The design of the SPTF is presented along with a comparison of its performance to calculated estimates of transmittance, spectral resolution, and spectral range.
KEYWORDS: Image segmentation, Hyperspectral imaging, Data modeling, Principal component analysis, Image processing, Imaging spectroscopy, Data processing, Data analysis
The information density in hyperspectral data is not uniform across the spectral and spatial dimensions, and the overall information sparsity is often high. While these non-uniformities underpin the sought-after image contrast, high sparsity generates unnecessarily long acquisition and data processing times. Conventional reduction techniques like those based on principal components analysis (PCA) sacrifice the contributions of minority pixel populations while retaining those representing a greater portion of the overall variability. The effect is that some regions in the reconstructed images achieve a higher degree of recovery than other locations, making it difficult to assess the meaning or relevance of the minority pixels, even when this information would reveal important sample defects or spectral inhomogeneities. In the work presented here, we introduce a novel user-unassisted data reduction and image segmentation method called reduction of spectral images (ROSI). The aim of ROSI is to achieve a threshold information density in the spectral dimension for all image pixels. The result effectively segments the image in a manner that provides rapid image contrast that is comparable to traditionally classified images, but does so without a priori information. In addition, ROSI results are suitable for subsequent data analysis and enable ROSI to be performed alone or as a preprocessing data reduction step. A full description of ROSI is presented along with results from both model and real hyperspectral data, and its performance is compared quantitatively to conventional class of data reduction methods.
KEYWORDS: Simulation of CCA and DLA aggregates, Chemical analysis, Imaging spectroscopy, Image enhancement, Image processing, Polymers, Principal component analysis, Image analysis, Data acquisition, Statistical analysis
Advances in spectral imaging instrumentation during the last two decades has lead to higher image fidelity, tighter spatial resolution, narrower spectral resolution, and improved signal to noise ratios. An important sub-classification of spectral imaging is chemical imaging, in which the sought-after information from the sample is its chemical composition. Consequently, chemical imaging can be thought of as a two-step process, spectral image acquisition and the subsequent processing of the spectral image data to generate chemically relevant image contrast. While chemical imaging systems that provide turnkey data acquisition are increasingly widespread, better strategies to analyze the vast datasets they produce are needed. The Generation of chemically relevant image contrast from spectral image data requires multivariate processing algorithms that can categorize spectra according to shape. Conventional chemometric techniques like inverse least squares, classical least squares, multiple linear regression, principle component regression, and multivariate curve resolution are effective for predicting the chemical composition of samples having known constituents, but are less effective when a priori information about the sample is unavailable. To address these problems, we have developed a fully automated non-parametric technique called spectral identity mapping (SIMS) that reduces the dependence of spectral image analysis on training datasets. The qualitative SIMS method provides enhanced spectral shape specificity and improved chemical image contrast. We present SIMS results of infrared spectral image data acquired from polymer coated paper substrates used in the manufacture of pressure sensitive adhesive tapes. In addition, we compare the SIMS results to results from spectral angle mapping (SAM) and cosine correlation analysis (CCA), two closely related techniques.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.