Open Access
Add to BibSonomyAbstract
Polarization is an important characteristic of electromagnetic waves, which can not be detected by either the human visual system or traditional image sensors. Motivated by various animal species with polarization vision as well as by the prospect of improving the image quality of the imaging systems, we are exploring the potential of polarization for microscope imaging. The most powerful techniques for molecule monitoring requires complex preprocessing for labeling the sample with different dyes. In this paper, we propose a cell detection method using polarization imaging without any need for staining target cell samples with any chemical dye. The motivation for this work is to develop an optical imaging technique that is simple and that can be used on live cells. The polarization sensitivity of cell samples is studied in this paper. A definition for the quantity called “polarization deviation” is proposed in order to identify clearer the difference between target cells and the background. Based on the polarization deviation detection method, a three-parameter polarization imaging method is employed to further simplify the image capture procedure for the proposed label-free cell detection. A color imaging methodology based on the well-known color space is utilized in order to represent the captured polarization information using computer graphics.
© 2012 Optical Society of America
1. Introduction
In biology, it is well known that cells isolated from different organs maintain their characteristics of original tissue, allowing researchers to examine the nature of live bodies. Among the methods to examine the cell characteristics, imaging is one of the most important and fundamental approaches for gathering the information and understanding the cell behavior, growth, morphology and its protein expressions. Comparing the other experimental approaches such as extracted protein, and RNA/DNA assays, imaging provides important information including expression levels and localization in fixed and live cells [1]. It is known that one of the challenges is how to detect a nearly transparent target specimen from a transparent background. In order to enhance the contrast between the target and the background under a microscope view, staining and fluorescence tagging [2–4] are widely used for cell monitoring. In this method of staining and tagging, cells are first fixed and stained by antibody-antigen reaction, followed by fluorescence labeling. The particular antigen (protein) expression can be detected by fluorescence microscopy. While this approach is powerful, the preparation of the target sample is time consuming, and cannot be utilized for live cell detection. In addition, highly sensitive and reliable antibodies that bind specifically to target protein/antigens of interest are needed for high accuracy. An alternative method is to manipulate the gene of the cells to express fluorescence to label cells or to detect the target gene expression, which can be detected under a fluorescence microscope. However, it is a costly and time consuming method. Any method for high accuracy cell detection is important for cell characterization, as well as for fast diagnosis of pathological conditions.
The invention of phase contrast microscopy [5, 6] has provided non-staining, live cell monitoring with high image quality. Phase contrast microscopy examines the phase shifts, which is invisible to the human visual system, of the incident light passing through a nearly transparent specimen. Inspired by the improvement of image quality via introducing the “invisible” phase of light into visible computer graphics, in this paper, we are introducing a new methodology of non-staining, live cell detection using another “invisible” characteristics of light, namely, polarization. The goal is to provide a simple but powerful method to help detect cells against its background that can be used on live cells. Polarization is the orientation of oscillation of the electric field vector (E vector) of a uniform plane light wave. Intensity, wavelength and polarization [7,8] are three of the main characteristics of visible light. However, human eyes, as well as traditional image sensors, are polarization-blind. Thus only two characteristics of visible light, intensity and wavelength, interpreted as perceptual qualities of brightness and color, are detected and encoded. On the other hand, according to the work of many biologists and zoologists in the past, the visual system of some animal species (e.g., honeybees, desert ants, mantis shrimps) can sense the polarization of light [9–12]. Polarization information can be employed to improve the image quality while sensing under optically scattered environments [13–15], such as underwater [16–19], or foggy/turbid conditions [20, 21]. Previous work also showed a great potential of polarization in remote sensing [22–25], fingerprint detection [26], sensing object illuminated by low intensity source light [27], or sensing through semireflection media [28]. In addition, as an important feature of electromagnetic waves, polarization of the electromagnetic waves carries potentially useful information, such as the surface orientation and material characteristic of the observed objects, which provides useful information for target detection [29–34].
In this paper, we explore a polarization-based methodology to enhance the detection sensitivity of conventional phase contrast cell detection using a conventional microscope without performing the time consuming and expensive cell staining using antibodies, chemical dyes or molecular tools to express fluorescence. In addition, this technique can be employed for live cell detection. We demonstrate in this paper that this approach, when added to the microscope-based imaging, can provide considerable enhancement in the cell detection without any need for special preparation of samples. Once the improved image has been obtained, various imaging algorithms (such as edge detection) can be applied. An active polarization system is employed using a linearly polarized artificial light to illuminate the target specimen. Transmitted light intensity is detected and analyzed. Quantitative results demonstrate considerable improvement in detection sensitivity of the proposed detection methodology when compared to intensity-only images.
2. Polarization-based cell detection
While observing cells under a microscope, the cells are defined as the target signal to be detected. A local region containing target cells is labeled as a “signal-present” region, while a zone without any target cells is defined as a“signal-absent” region. The question is how to distinguish the “signal-present” region from the “signal-absent” region within the microscope ocular scene. Human pulmonary vascular smooth muscle cells are employed for this experimental testing. The target cells are first seeded and cultured on the chamber slides in the CO2 incubator (37°C with 5% CO2), and then fixed with 4% paraformaldehyde for protein/ morphology preservation followed by washing with phosphate buffered saline (PBS). Fixed cells are kept in PBS at 4°C and their morphology is examined under an uplight microscope (Nikon Eclipse TE2000-U). Figure 1(a) illustrates the image captured under unpolarized illumination. Two subregions are zoomed in and labeled as Region I and Region II in Fig. 1(b), respectively.
Fig. 1 (a) Raw image captured under unpolarized illumination. (b) A zoom-in view of local zone, two subregions are labeled as Region I and Region II, respectively.
We first examine the polarization sensitivity of the cell samples. A Nikon Eclipse TE2000-U, which is an upright microscope, is utilized for image acquisition. A one-inch-diameter glass polarizer is inserted horizontally between the light source and the sample in order to generate a linearly polarized illumination. We collect 18 intensity images using polarization-blind image acquisition system. The images are captured while the sample is illuminated by a linearly polarized incident light with different orientation angles of its linear polarization vector. The angle χ varies from 0° to 180° in steps of 10°. Within a local region of the sample as shown in Fig. 1(b), the digitized intensity of some randomly selected pixels from the two subregions Region I and Region II captured under illumination of incident light with different polarization angles are plotted in Fig. 2(a).
Fig. 2 (a) The digitized value of randomly selected pixels captured under different polarization angles of the linearly polarized illumination light, from Region I (as illustrated in red) and Region II (as illustrated in blue), respectively. (b) The distribution of the polarization Deviation σp for Region I (as illustrated in red) and Region II (as illustrated in blue), respectively.
The variation of the captured intensity for the same pixel but under different polarization angles of the illumination is considerably larger in Region I than in Region II. In order to quantitatively evaluate the variation of the captured intensity for the same pixel but under different illumination polarization, we define the standard deviation of the captured intensity values of one pixel located at coordinates (x,y) illuminated by linearly polarized light with an angle of polarization χi(i = 0, 1...(n − 1)) as “Polarization Deviation”, σp(x, y),
where μp(x, y) represents a “Polarization Mean” defined as Figure 2(b) illustrates the distribution of σp of the two regions. It is striking that the variation σp of the majority of the pixels in Region I (where the cell is present) is considerably larger than in Region II (where no cell is present). This provides a way to identify regions where cells are present, i.e. where the deviation σp is larger than a threshold σth. On the other hand, a region where the majority of the pixels has a deviation smaller than a threshold σth indicates that no cells are present.The raw captured unpolarized image is illustrated in Fig. 3(a). An enhanced image after histogram equalization [35] are illustrated in Fig. 3(b). The value of σp is further scaled up into the full range of a gray color map as represented in Fig. 3(c), which is denoted as a “Polarization Deviation Image”. Comparing the polarization deviation image with the raw image captured under unpolarized illumination and an enhanced raw image after histogram equalization, the polarization deviation image appears to provide a “better” and more distinguishable image as illustrated in Fig. 3. In order to quantitatively evaluate the capability of distinguishing the “signal-present” pixels from “signal-absent” pixels, a sensitivity index d′ is defined as
where μs and μn are the mean of pixel values of polarization deviation from “signal-present” region and “signal-absent” region, respectively, while and are the variance of pixel values of such polarization deviations from the two regions. For the raw captured and enhanced image, “pixel value” represents light intensity, while for the polarization Deviation image, it means the polarization Deviation strength. The sensitivity index of the raw image captured under unpolarized illumination, an enhanced raw image after histogram equalization and the polarization Deviation image are listed under the images in Fig. 3. The results clearly show that the sensitivity is improved considerably by introducing the polarization Deviation.Fig. 3 (a) Raw captured intensity image under unpolarized illumination; (b) enhanced intensity image after histogram equalization, and (c) the image formed by distribution of the polarization deviation. The sensitivity index d′ is listed under each image.
Based on the polarization Deviation value, one pixel is categorized as a “signal-present” pixel when the polarization Deviation value is larger than a threshold σth, selected by the operator. Thus, effectively a threshold filter is introduced to the detection system.
As illustrated in Fig. 4. The pixel categorized as “signal-present” pixels are labeled as white and those as “signal-absent” are labeled as black pixels. In order to suggest a design guideline for the polarization cell detection methodology, we optimize the threshold value of the polarization deviation with respect to the optimum detection sensitivity. Various “signal-present” subregions are selected as illustrated in Fig. 5(a). The sensitivity index of each sample regions changes as the threshold σth is increased as plotted in Fig. 5(b). It should be noted that although an optimum threshold value varies for different samples, we found that a threshold around 30 provides a good overall detection sensitivity.Fig. 4 Categorization results of “signal present” and “signal absent” employing a σth equals to (a) 30; (b) 40; and (c) 50.
Fig. 5 A comparison of sensitivity index of different standard Deviation threshold values. (a) various “signal-present” regions are selected as testing samples; (b) The sensitivity index as shown on the y-axis changes as the threshold σth is increased.
3. Three-parameter polarization-based cell detection
The results of the previous section show that the polarization-based cell detection capability is a promising technique. However, a large amount of raw images are still required to be captured under different polarization angles of the linearly polarized illuminating light. In this section, we propose to perform polarization-based cell detection using a three-parameter polarization imaging method.
If we observe the captured intensity from a single pixel, located in “signal present” region, under illuminations with different polarization angles as illustrated in Fig. 6, the variation of the captured intensity for the pixel generally posses a sinusoidal shape. Therefore, we assume the intensity distribution can be represented by a sinusoidal fitting curve defined as
where χ is the orientation angle of the polarization vector of the incident polarized wave. As can be seen, there are three parameters involved in this sinusoidal variation intensity: average energy intensity L̄, the degree of intensity modulation p (or modulating intensity Lp), and the angle ψ at which the maximum intensity is detected. Therefore, the intensity values collected from three independent angles of the polarization of the incident wave are sufficient in order to predict the entire sinusoidal variation of the intensity of each pixel.Fig. 6 Value of the captured intensity of pixels under different angles of linearly polarized light. The pixels, represented by different colors, are randomly selected from the “signal present” region.
The combination of different angles χ of the incident linear polarization may in general be arbitrarily. Theoretically [36], an optimal arrangement of photoreceptors/photodetectors for a three-channel polarization visual/imaging system (PVS) is an arrangement that is equally angularly spaced over a 180° interval, i.e. χ1 = 0°, χ2 = 60°, χ3 = 120°.
Thus, the three unknown parameters, L̄, p (or Lp), and ψ can be extracted from solving the above equations as Where χ varies between to , and Lp = L̄ ·p, with a p varies between 0 to 1. p = 0 corresponds to the case of unvarying intensity in light detection, while p = 1 represents completely varying light detection as the angle of incident polarization varies. Previous studies [37–41] have clearly shown the potential enhancement of image quality by introducing polarization into traditional digital image processing. In the following, we will perform this three-parameter polarization imaging method on cell detection to illustrate the improvement of detection sensitivity.Figure 7 shows the microphotographies of the samples taken under different illumination condition of polarized light. The differences between each frame can not be distinguished by the naked eye. The degree of linear polarization p and the orientation angle χ can be derived from these three frames using Eq. (7). The p and Ψ values are linearly mapped to a gray level as illustrated in Fig. 8.
Fig. 7 Microphotography of the human pulmonary vascular smooth muscle cells illuminated under three different linear polarization angles of incident beam (a) χ1 = 0°, (b) χ2 = 60°, (c) χ3 = 120°.
Fig. 8 (a) The degree of linear polarization p which varies from 0 to 1; (b) the orientation angle Ψ calculated varying from to .
Random noise is introduced during the three-frame image capturing procedure. A single threshold filter as defined in Eq. (4) is employed in this three-parameter polarization imaging processing in order to reduce the effect of noise. The polarization Deviation σp(x, y) value is calculated based on the intensity captured from only three different polarization angles instead of 18 angles. Figure 9 shows a zoom-in view of scaled polarization degree (p) in the same local area shown in the previous section. Table 1 quantitatively compares the detection sensitivity of the raw captured image illuminated by unpolarized light, the standard Deviation image derived from a set of images captured under different polarized lights, and the polarization degree calculated from the three-parameter polarization imaging method using three frames illuminated by different polarized light. It should be noted that the standard Deviation image provides a higher detection sensitivity. However, under the same integration time for each frame during image acquisition, 18 frames of raw images are employed in polarization deviation imaging, comparing to only 3 frames of such images for the three-parameter polarization imaging. The amount of input data is decreased by 83.33% for the three-parameter method as compared to the polarization deviation imagining method, while the detection sensitivity is only reduced by 25.8%.
Fig. 9 The post-processing results of the polarization degree calculated from three-parameter polarization imaging, using a threshold filter employing different cut-off threshold value σth equals to (a) 30; (b) 40; and (c) 50.
Table 1. A Comparison of Detection Sensitivity of the Unpolarizedly Illuminated Image (I), the Standard Deviation Image (σp), and the three-parameter polarization image (Lp)
At the beginning of this section, in order to predict the entire sinusoidal variation of the intensity of each pixel, the combination of different polarization angles χ1 = 0°, χ2 = 60°, χ3 = 120° was selected according to Bernard and Wehner’s theoretical analysis [36]. Table 2, compares the detection sensitivity of the predicted sinusoidal variation Lp solved from Eq. (7) employing different polarization angle combinations. The detection sensitivity is tested on different sample regions as used above. A mean detection sensitivity of the various sample regions is illustrated. An angle combination with a Δχ around 60° does provide a higher detection sensitivity.
Table 2. A Comparison of Detection Sensitivity while Using Different Combination of Angles
In order to better present the polarization information, all the three parameters, L̄, p and Ψ, solved from Eq. (7) as a prediction of the sinusoidal distribution for each pixel are combined into one single frame. The human visual system is polarization-blind, but can well detect and encode intensity and color information. Inspired by the work of Tyo et al. [39] and Yemelyanov et al. [42], polarization information can be represented in a three-parameter polarization color vision system. The three parameters solved from Eq. (7) are illustrated in Fig. 10 using a well-known color 3-dimensional space (HSV domain) used in computer graphics by the parallelism of L̄ with the lightness/value (V), p with the saturation (S), and Ψ with the hue (H). Both the three-parameter polarization imaging integrated with/without a threshold filter are illustrated in Fig. 10. The high-pass filter greatly helps reducing the background noise from the target cells.
Fig. 10 Three-parameter polarization color plotting while (a) no filter is employed, and a high-pass filter with a cut-off threshold equals to (b) 30, (c) 40 and (d) 50 is used, respectively.
4. Polarization-based live cell detection
The proposed polarization-based cell detection method has been tested on live cell samples as illustrated in Fig. 11. The plotting of intensity of live cell samples captured under unpolarized illumination is compared with the pseudo HSV color plotting based on the three-parameter polarization methodology without/with a high pass filter. The first cell sample as illustrated in Fig. 11(a), is human pulmonary vascular smooth muscle cells (VSMCs), which were cultured on 10 cm culture plates placed in 5% CO2 humidified cell culture incubator at 37°C. The second specimen is human oral cavity epithelial cells as shown in Fig. 11(b). We sample the oral cavity epithelial cells by swiping the surface of mouth cavity by cotton swab. The cotton swab was smeared on a glass slide. A few drops of phosphate buffered saline (PBS) is added before covering the sample by glass for examination under the microscope.
Fig. 11 Live cell detection under unpolarized illumination (first row), and using the three-parameter polarization methodology without (second row) or with (third row) a high pass filter. The results from the three-parameter polarization detection method are plotted in HSV pseudo color domain. The sample cells include (a) human pulmonary vascular smooth muscle cells, and (b) human oral epithelial cavity cells.
Different parts of the cell can be represented by a different polarization degree. The proposed methodology improves the cell detection sensitivity under standard microscope without the requirement of the time consuming preparation and expensive immune staining analysis. The proposed polarization-based cell detection works on live cells as well, which shows a potential to develop convenient and low cost imaging for cell characterization for diagnosis and experimental detections.
5. Conclusion
Cell imaging, as an important approach in the cell behavior studies, provides important information about the target specimens, such as expression levels. Common cell detection approaches, including fluorescence staining and gene expression, enable high quality imaging, but are time consuming. In addition, those approaches are not available for live cell imaging. In this paper, we have explored the potential applications of polarization-based cell detection. According to experimental results, the cell sample is sensitive to the change of polarization angles of the illuminating light. A polarization deviation methodology is derived in order to evaluate the sensitivity of the cell samples to different polarization illuminations. Experimental results clearly show that the detection sensitivity can be improved by introducing polarization. Based on the polarization Deviation method, we have further simplified the polarization based cell detection methodology into a three-parameter cell detection method which requires intensity captured from only three independent polarization angles of linearly polarized illuminating light. The proposed detection methodology has been performed on non-staining, monolayer, live cells, demonstrating the usefulness of our technique.
Acknowledgment
This work is supported in part by National Science Foundation, through grants EFRI-1038215. We thank T. Nguyen, from the department of pathology and laboratory medicine of University of Pennsylvania, for his effort in taking care of the cell sample. We also thank A. Higgins, an undergraduate candidate in the department of system engineering of University of Pennsylvania, for her effort in data collection.
References and links
1. D. Stephens, Cell Imaging (Scion, 2006).
2. H. J. Conn, Biological Stains: A Handbook on the Nature and Uses of the Dyes Employed in the Biological Laboratory (Biotech Publications, 1953).
3. G. Clark, Staining Procedures (Williams and Wilkins, 1981).
4. G. Clark, F. H. Kasten, and H. J. Conn, History of Staining (Williams and Wilkins, 1981).
5. F. Zernike, “Phase contrast, a new method for the microscopic observation of transparent objects,” Physica 9, 686–698 (1942). [CrossRef]
6. F. Zernike, “Phasecontrast, a new method for the microscopicobservation of transparentobjects part II,” Physica 9, 974–980 (1942). [CrossRef]
7. D. Goldstein, “Polarized Light” (Marcel Dekker: New York, 2003). [CrossRef]
8. T. H. Waterman, “Polarization sensitivity” in The Handbook of Sensory Physiology, vol. VII/6B Vision in Invertebrates, edited by H. Autrum, (Springer-Verlag, New York, 1981). [CrossRef]
9. R. Wehner, “Polarized-light navigation by insects,” SCI AM 235, 106–114, (1976). [CrossRef] [PubMed]
10. T. Labhart, “Polarization opponent interneurons in the insect visual system,” Nature 331, 435–437 (1988). [CrossRef]
11. R. Wehner, “Neurobiology of polarization vision,” TRENDS NEUROSCI 12, 353–359, (1989). [CrossRef] [PubMed]
12. T. W. Cronin and J. Marshall, “Parallel processing and image analysis in the eyes of mantis shrimps,” Biol. Bull. 200, 177–183 (2001). [CrossRef] [PubMed]
13. S. G. Demos, W. B. Wang, and R. R. Alfano, “Imaging objects hidden in scattering media with fluorescence polarization preservation of contrast agents,” Appl. Opt. 37, 792–797 (1998). [CrossRef]
14. J. G. Walker, P. C. Chang, and K. I. Hopcraft, “Visibility depth improvement in active polarization imaging in scattering media,” Appl. Opt. 39, 4933–4941 (2000). [CrossRef]
15. H. H. Wang, C. W. Sun, Y. M. Wang, Y. W. Kiang, and C. C. Yang, “Determination of the depth of a scattering target in a turbid medium with polarization discrimination of transmitted signals,” Opt. Lett. 28, 25–27 (2003). [CrossRef] [PubMed]
16. P. C. Y. Chang, J. C. Flitton, K. I. Hopcraft, E. Jakeman, D. L. Jordan, and J. G. Walker, “Improving visibility depth in passive underwater imaging by use of polarization,” Appl. Opt. 42, 2794–803 (2003). [CrossRef] [PubMed]
17. Y. Y. Schechner, N. Karpel, T. Israel, and I. Technology, “Clear Underwater Vision,” Pattern Recogn 1, 536–543 (2004).
18. Y. Y. Schechner and N. Karpel, “Recovery of Underwater Visibility and Structure by Polarization Analysis,” IEEE J Oceanic Eng. 30, 570–587 (2005). [CrossRef]
19. T. Treibitz and Y. Y. Schechner, “Active polarization descattering,” IEEE Trans. Pattern Anal. Mach. Intell. 31, 385–399 (2009). [CrossRef] [PubMed]
20. S. Shwartz, E. Namer, and Y. Schechner, “Blind Haze Seperation,” in Proceedings of IEEE Computer Society Conference on Computer Vision and Pattern Recognition (IEEE2006), pp. 1984–1991.
21. J. S. Tyo, M. P. Rowe, E. N. Pugh Jr., and N. Engheta, “Target detection in optically scatter media by polarization-difference imaging”, Appl. Opt. 35, 1855–1870 (1996). [CrossRef] [PubMed]
22. B. Lyot, Research on the Polarization of Light from Planets and Some Substances (Ann. Observatoire de Paris, Section de Meudon1929).
23. W. G. Egan, W. R. Johnson, and V. S. Whitehead, “Terrestrial polarization imagery obtained from the Space Shuttle: characterization and interpretation,” Appl. Opt. 30, 435–42 (1991). [CrossRef] [PubMed]
24. J. L. Deuz, F. M. Bron, C. Devaux, P. Goloub, M. Herman, B. Lafrance, F. Maignan, A. Marchand, F. Nadal, G. Perry, and D. Tanr, “Remote sensing of aerosols over land surfaces from POLDER-ADEOS-1 polarized measurements”, J. Geophys. Res. 106, 4913–4926 (2001). [CrossRef]
25. J. S. Tyo, D. L. Goldstein, D. B. Chenault, and J. A. Shaw, “Review of passive imaging polarimetry for remote sensing applications,” Appl. Opt. 45, 5453–5469 (2006). [CrossRef] [PubMed]
26. S. S. Lin, K. M. Yemelyanov, E. N. Pugh Jr., and N. Engheta, “Polarization-based and specular-reflection-based noncontact latent fingerprint imaging and lifting,” J. Opt. Soc. Am. A 23, 2137–2153 (2006). [CrossRef]
27. S. S. Lin, K. M. Yemelyanov, E. N. Pugh Jr., and N. Engheta, “Separation and contrast enhancement of overlapping cast shadow components using polarization,” Opt. Express 14, 7099–7108 (2006). [CrossRef] [PubMed]
28. Y. Y. Schechner, J. Shamir, and N. Kiryati, “Vision through semireflecting media: polarization analysis,” Opt. Lett. 24, 1088–1090 (1999). [CrossRef]
29. T. Germer and M. Fasolka, “Characterizing surface roughness of thin films by polarized light,” Proc. SPIE 518, 264–275 (2003). [CrossRef]
30. D. Miyazaki, R. Tan, K. Hara, and K. Ikeuchi, “Polarization-based Inverse Rendering from a Single View,” in Proceedings of Ninth IEEE International Conference on Computer Vision (IEEE2003), pp. 982–987. [CrossRef]
31. F. Goudail and P. Rfrgier, “Statistical algorithms for target detection in coherent active polarimetric images,” J. Opt. Soc. Am. A 18, 3049–3060 (2001). [CrossRef]
32. F. Goudail and P. Rfrgier, “Statistical techniques for target detection in polarization diversity images,” Opt. Lett. 26, 644–646 (2001). [CrossRef]
33. F. Goudail, P. Terrier, Y. Takakura, L. Bigu, F. Galland, and V. DeVlaminck, “Target detection with a liquid-crystal-based passive Stokes polarimeter.,” Appl. Opt. 43, 274–82 (2004). [CrossRef] [PubMed]
34. R. Nothdurft and G. Yao, “Expression of target optical properties in subsurface polarization-gated imaging,” Opt. Express 13, 4185–4195 (2005). [CrossRef] [PubMed]
35. R. C. Gonzalez and P. Wintz, “Image Enhancement,” in Digital Image Processing (Addison Weslwy, 1987), pp. 146–152.
36. G. D. Bernard and R. Wehner, “Functional similarities between polarization vision and color vision,” Vision Res. 17, 1019–1028 (1977). [CrossRef] [PubMed]
37. M. P. Rowe, E. N. Pugh Jr., and N. Engheta, “Polarizationdifference imaging: a biologically inspired technique for observation through scattering media,” Opt. Lett. 20, 608–610 (1995). [CrossRef] [PubMed]
38. J. S. Tyo, M. P. Rowe, E. N. Pugh Jr., and N. Engheta, “Target detection in optically scattering media by polarization-difference imaging,” Appl. Opt. 35, 1855–1870 (1996). [CrossRef] [PubMed]
39. J. S. Tyo, E. N. Pugh Jr., and N. Engheta, “Colorimetric representation for use with polarization-difference imaging of objects in scattering media,” J. Opt. Soc. Am. A 15, 367–374 (1998). [CrossRef]
40. S.-S. Lin, K. M. Yemelyanov, E. N. Pugh Jr., and N. Engheta, “Separation and contrast enhancement of overlapping cast shadow components using polarization,” Opt. Express 14, 7099–7108 (2006). [CrossRef] [PubMed]
41. S.-S. Lin, K. M. Yemelyanov, E. N. Pugh Jr., and N. Engheta, “Polarization-based and specular-reflection-based noncontact latent fingerprint imaging and lifting,” J. Opt. Soc. Am. A 23, 2137–2153 (2006). [CrossRef]
42. K. M. Yemelyanov, S. S. Lin, W. Q. Luis, E. N. Pugh Jr., and N. Engheta, “Bio-inspired display of polarization information using selected visual cues,” Proc. of SPIE 5158, 71–84 (2003). [CrossRef]
