OSA's Digital Library

Virtual Journal for Biomedical Optics

Virtual Journal for Biomedical Optics

| EXPLORING THE INTERFACE OF LIGHT AND BIOMEDICINE

  • Editors: Andrew Dunn and Anthony Durkin
  • Vol. 7, Iss. 7 — Jun. 25, 2012
« Show journal navigation

Optical turbulence on underwater image degradation in natural environments

Weilin Hou, Sarah Woods, Ewa Jarosz, Wesley Goode, and Alan Weidemann  »View Author Affiliations


Applied Optics, Vol. 51, Issue 14, pp. 2678-2686 (2012)
http://dx.doi.org/10.1364/AO.51.002678


View Full Text Article

Acrobat PDF (586 KB)





Browse Journals / Lookup Meetings

Browse by Journal and Year


   


Lookup Conference Papers

Close Browse Journals / Lookup Meetings

Article Tools

Share
Citations

Abstract

It is a well-known fact that the major degradation source on electro-optical imaging underwater is from scattering by particles of various origins and sizes. Recent research indicates that, under certain conditions, the apparent degradation could also be caused by the variations of index of refraction associated with temperature and salinity microstructures in the ocean and lakes. The combined impact has been modeled previously through the simple underwater imaging model. The current study presents the first attempts in quantifying the level of image degradation due to optical turbulence in natural waters in terms of modulation transfer functions using measured turbulence dissipation rates. Image data collected from natural environments during the Skaneateles Optical Turbulence Exercise are presented. Accurate assessments of the turbulence conditions are critical to the model validation and were measured by two instruments to ensure consistency and accuracy. Optical properties of the water column in the field were also measured in coordination with temperature, conductivity, and depth. The results show that optical turbulence degrades the image quality as predicted and on a level comparable to that caused by the particle scattering just above the thermocline. Other contributing elements involving model closure, including temporal and spatial measurement scale differences among sensors and mitigation efforts, are discussed.

1. Background

When it comes to diver visibility, the most prominent issue is the degradation of image quality over distance due to signal attenuation. This presents a striking contrast for those who are used to the seemingly unlimited visual ranges in air. The cause of the degradation has been mostly attributed to the “dirt,” or inorganic as well as organic particles (i.e., microorganisms and detritus) in the water, and rightly so. Most research has focused on reducing the impact of particle scattering by means of discriminating scattering photons involving polarization, range gating, modulation, and by means of restoration via deconvolution [1

1. G. D. Gilbert and J. C. Pernicka, “Improvement of underwater visibility by reduction of backscatter with a circular polarization technique,” Appl. Opt. 6, 741–746 (1967). [CrossRef]

6

6. G. R. Fournier, D. Bonnier, J. L. Forand, and P. W. Pace, “Range-gated underwater laser imaging system,” Opt. Eng. 32, 2185–2190 (1993). [CrossRef]

]. However, in clean oceanic or lake waters, another factor could come into play in effects of the same order. This is the scattering by optical turbulence, which is caused by variations of the index of refraction of the medium. This is mostly associated with the turbulence structures of the medium, or the water body in our case. Degradation of the image quality in a scattering medium involving turbulence has been studied mostly in atmosphere [7

7. D. L. Fried, “Optical resolution through a randomly inhomogeneous medium for a very long and very short exposures,” J. Opt. Soc. Am. 56, 1372–1379 (1966). [CrossRef]

9

9. N. Kopeika, “Imaging through the atmosphere for airborne reconnaissance,” Opt. Eng. 26, 1146–1154 (1987). [CrossRef]

]. These studies are mainly focused on modeling the optical transfer function as a function of density variations by association with wind profiles in an effort to restore the images obtained, such as in air reconnaissance and astronomy studies [10

10. D. Sadot, A. Dvir, I. Bergel, and N. Kopeika, “Restoration of thermal images distorted by the atmosphere, based on measured and theoretical atmospheric modulation transfer function,” Opt. Eng. 33, 44–53 (1994). [CrossRef]

,11

11. Y. Yitzhaky, I. Dror, and N. Kopeika, “Restoration of atmospherically blurred images according to weather-predicted atmospheric modulation transfer functions,” Opt. Eng. 36, 3064–3072 (1997). [CrossRef]

]. Little has been done regarding the turbulence effects on imaging formation in water, mainly due to the dominant particle scattering and associated attenuation. This is of little surprise to anyone with experience in coastal waters, especially those inside a harbor, or estuarine areas like that of the Mississippi, where visibility could quickly reduce to zero in a matter of a few feet. The same applies to regions of strong resuspension from the bottom, both in coastal regions as well as in the deep sea. The effects of turbulence have been postulated to have impacts only over long image-transmission ranges [12

12. W. H. Wells, “Theory of small angle scattering,” in AGARD Lecture Series (NATO, 1973), Vol. 61.

], which has been supported by light-scattering measurements and simulations [13

13. D. J. Bogucki, J. A. Domaradzki, R. E. Ecke, and C. R. Truman, “Light scattering on oceanic turbulence,” Appl. Opt. 43, 5662–5668 (2004). [CrossRef]

]. Under extreme conditions, observations have been made that involve targets with a path length of a few feet [14

14. G. D. Gilbert and R. C. Honey, “Optical turbulence in the sea,” in Underwater Photo-optical Instrumentation Applications (SPIE, 1972), pp. 49–55.

]. The images obtained under such conditions are often severely degraded or blurred, on par with or more degraded than those images where blurring is caused by particle scattering. Overcoming such challenges to increase both the reach and the resolution is important to current and future underwater electro-optical (EO) applications. It is critical to establish a good understanding about the limiting factors under different conditions. The simple underwater imaging model (SUIM) [15

15. W. Hou, “A simple underwater imaging model,” Opt. Lett. 34, 2688–2690 (2009). [CrossRef]

] was developed to address this issue, and it has been shown in theory that, on average, the relative contribution of different components [3

3. W. Hou, Z. Lee, and A. Weidemann, “Why does the Secchi disk disappear? An imaging perspective,” Opt. Express 15, 2791–2802 (2007). [CrossRef]

,16

16. W. Hou, D. Gray, A. Weidemann, and R. Arnone, “Comparison and validation of point spread models for imaging in natural waters,” Opt. Express 16, 9958–9965 (2008). [CrossRef]

] in underwater imaging applications can be expressed in terms of the optical transfer function (OTF) as
OTF(ψ,r)total=OTF(ψ,r)pathOTF(ψ,r)parOTF(ψ,r)tur=(11+D)exp[cr+br(1e2πθ0ψ2πθ0ψ)]exp(Snψ5/3r)=(11+D)exp{[cb(1e2πθ0ψ2πθ0ψ)+Snψ5/3]r},
(1)
where θ0 relates to the mean scattering angle and c and b are the beam attenuation and scattering coefficients, respectively. ψ is the spatial frequency in cycles per radian, r is the imaging range, and D relates to path radiance [3

3. W. Hou, Z. Lee, and A. Weidemann, “Why does the Secchi disk disappear? An imaging perspective,” Opt. Express 15, 2791–2802 (2007). [CrossRef]

]. Sn contains parameters that are dependent on the structure function, which can be further expressed in terms of the turbulence dissipation rate of temperature, salinity, and kinetic energy, assuming a Kolmogorov power spectrum type:
Sn=3.44(λ¯/R0)5/3=1736K3λ¯1/3,
(2)
where K3=B1χε1/3 and reflects the three-dimensional (3D) optical turbulence strength. λ¯ is the average wavelength of the transmitted light. R0 is the characteristic seeing parameter [15

15. W. Hou, “A simple underwater imaging model,” Opt. Lett. 34, 2688–2690 (2009). [CrossRef]

]. B1 is a constant and is assumed to be on the order of unity [17

17. G. K. Batchelor, “Small-scale variation of convected quantities like temperature in turbulence fluid,” J. Fluid Mech. 5, 113–133 (1959). [CrossRef]

]. ε is the turbulence kinetic energy dissipation (TKED) rate. χ relates to the dissipation rate of temperature (TD) or salinity variances [17

17. G. K. Batchelor, “Small-scale variation of convected quantities like temperature in turbulence fluid,” J. Fluid Mech. 5, 113–133 (1959). [CrossRef]

]. For convenience, key parameters used are listed in Table 1.

Table 1. Key Parameters and Definitions Used in the Paper

table-icon
View This Table

2. SOTEX

A. Experiment Site

Optical turbulence underwater is primarily a function of temperature structure, although salinity variations could at times contribute to strong optical turbulence [14

14. G. D. Gilbert and R. C. Honey, “Optical turbulence in the sea,” in Underwater Photo-optical Instrumentation Applications (SPIE, 1972), pp. 49–55.

]. Intensified thermoclines in the natural environment provide a convenient setup to examine this stochastic process. We identified one of the Finger Lakes in upstate New York, Skaneateles, as our test site for the Skaneateles Optical Turbulence Exercise (SOTEX, July 2010). Figure 1 shows the approximate location of the two sampling stations. The first (S1, red circle) is near the center of the lake (42.8668° N, 76.3920° W) over a sloping bottom with an approximate depth of 70 m, while the second (S2, blue triangle) is at the northern end of the lake (42.9063° N, 76.4058° W) over a flatter bottom with an approximate depth of 50 m. The lake is the clearest of all the Finger Lakes, with an average Secchi depth around 8 m [18

18. S. W. Effler, A. R. Prestigiacomo, and D. M. O’Donnel, “Water quality and limnological monitoring for Skaneateles Lake: field year 2007” (Upstate Freshwater Institute, 2008), p. 57.

], and allows for imaging under varied turbulent strength with little scattering contribution from particulates. The strong stratification in July at a relatively shallow depth, with low wind and low current interference, ensures a well-defined thermocline, as demonstrated by the temperature profile shown in Fig. 2, measured by a Precision Measurements Engineering (PME) conductivity and temperature (CT) sensor. The same features help to form strong structures for optical turbulence in the lake. The optical properties of the water column were measured by a nine-channel absorption and attenuation meter (ac-9, WETLabs), and laser in situ scattering transmissometer (LISST; Sequoia Scientific). We notice that the top part of the water column, from the surface to a depth of about 9 m, just above the sharp thermocline, is typically rather clear, with beam attenuation at 532 nm valued below 0.45m1 (including water absorption), then increasing sharply with increasing depth for the day shown (July 27, Fig. 2). The LISST data are primarily used to estimate the point spread function (PSF) of the water, which can be used for image model restoration [16

16. W. Hou, D. Gray, A. Weidemann, and R. Arnone, “Comparison and validation of point spread models for imaging in natural waters,” Opt. Express 16, 9958–9965 (2008). [CrossRef]

] as well as estimation of MTF. This effort will be discussed in a separate publication.

Fig. 1. Bathymetric sketch of Skaneateles Lake showing the approximate location of the two stations: S1 (red circle) near the center of the lake, and S2 (blue triangle) in the northern end of the lake. Map from http://www.ourlake.org/html/skaneateles_lake1.html.
Fig. 2. Optical properties (beam-c at 532 nm) and temperature profile measured during July 27 daytime IMAST deployment. Temperature profiles from other deployments are plotted as well to show the stable and strong thermoclines at station 1.

B. Instrumentation and Measurements

Fig. 3. IMAST night deployment configuration: imaging camera and housing (left end of the frame); active target made out of iPad (right end of the frame). Notice the Vector and CT locations. For details on other sensors, including the Vector, please refer to the text.

The SUIM model requires accurate assessments of turbulence structures of both the rates of TKED, ε, and TD, χ. Two sets of instrumentation were utilized for characterizing the turbulence strength during SOTEX in order to provide a comparison between the background turbulence of the lake and the turbulence within the IMAST, to ensure the IMAST itself was not inducing significant turbulence. This setup was necessary as we wanted to have optical turbulence measured during the imaging process. The first instrumental setup consisted of a Vector and a PME CT. The two instruments were mounted near the center of the IMAST structure, and the heads of the instruments placed in such a way as to sample the same volume of water, thus providing a time series of the 3D velocity, temperature, and conductivity fluctuations of the sample water volume (Fig. 3). As the instrument is commonly used for laboratory measurements or stationary moorings, the instrument requires collection of a time series of velocities at a stationary depth in order to compute the turbulent dissipation rates. Therefore, during deployment, the IMAST profiled the water column by pausing at each depth for 5 to 15 min to capture the turbulence statistics. This also helped to reduce the artificial turbulence induced by the IMAST, allowing for sampling under conditions closer to those of the natural environment. The second turbulence instrument, providing microstructure observations for SOTEX, was the Rockland Scientific Vertical Microstructre Profiler (VMP). It is equipped with four microstructure sensors: two shear sensor probes, one thermistor (FP07), and a microconductivity (SBE7) sensor. These sensors allow measurement of the microscale velocity shear, temperature, and conductivity with high accuracy and resolution. Additionally, the VMP profiler had externally attached SeaBird temperature and conductivity sensors.

The free-falling deployment nature of the VMP is not suitable for IMAST deployment needs despite its high temporal sampling rate and technological maturity in quantifying underwater turbulence. Rather, the compactness of the Vector/CT setup was ideal for our application. However, it is not designed for turbulence profiling; therefore a comparison or an intercalibration was necessary. The abovementioned dual setup helped to ensure the data quality and integrity. Details of the turbulence measurement methods and results may be found in [19

19. S. Woods, Naval Research Laboratory, Stennis Space Center, MS 39529, USA, W. Hou, E. Jarosz, W. Goode, and A. Weidemann are preparing a manuscript be called “Measurements of turbulence dissipation in Lake Skeneateles.”

].

To keep the IMAST to a manageable size and weight for ease of deployment, the imaging system, the target, and the environmental characterization sensors were all made to be self-contained. Besides a reflective custom-made resolution chart set, an active target was made utilizing an iPad for active illumination, displaying a standard USAF 1951 chart in order to reduce the effects of path radiance. A commercial off-the-shelf (COTS) high-speed imaging camera by Casio (EX-F1) was used to record images at various frame rates up to 300 fps at 500×400 pixels, and higher resolution with reduced frame rates including native full high-definition (HD) resolution (1920×1080 pixels). Because only the relative contrasts of different spatial frequencies are needed, we ignore the irradiance variations from the targets or exposure settings on the camera so long as they remain fixed and constant during each cast. IMAST was deployed in two configurations in the water column, vertical and horizontal. In the horizontal mode, the frame was suspended in the water column in the middle of, above, and beneath the thermocline for extended periods of time in order to accurately quantify the turbulence environment (Fig. 4).

Fig. 4. Diagram of deployment setup showing alternate deployment configurations: Vector/CT deployed on IMAST both vertically and horizontally. Note, in both instances, the VMP was deployed from a separate vessel. The intercomparison is necessary in order to measure turbulence impacts during IMAST deployments.

3. Results and Model Validation

The impact of the optical turbulence can be best illustrated when a pair of sample images is examined side by side [Figs. 5(a) and 5(b), cropped from full frame and enlarged to show details], one with strong optical turbulence and one without, under similar turbidity conditions. They are taken at two different depths: one at 2.8 m, which is essentially free of optical turbulence, while the one at 8.7 m is strongly influenced by optical turbulence. Measurements at other depths do not add extra information, especially considering the short span of the thermocline. Therefore they are not presented in this contrast study. From the temperature profiles shown in Fig. 2, one should expect a negligible amount of optical turbulence at the shallow depths where the temperature profile is essentially uniform, as reflected in very low TD rates. Both images are taken under the horizontal deployment configuration. One may notice that despite the similarity in measured optical properties (beam c less than 0.45m1, Fig. 2), the image taken inside the strong turbulence layer [Fig. 5(b)] suffered much more degradation compared to that obtained under conditions of weaker turbulence [Fig. 5(a)]. It is worth mentioning that for the IMAST iPad setting, the 0-2 group of the USAF 1951 resolution chart corresponds to the spatial frequency of 1900cycles/rad or 1.9cycles/mrad.

Fig. 5. Sample image pair obtained by IMAST during night deployment (IMAST horizontal) of July 27. The corresponding physical conditions can be seen in Fig. 2 and related publications (see text). (a) The left was taken at 2.8 m depth with no obvious optical turbulence, while (b) the right was from 8.7 m under conditions of similar turbidity but strong optical turbulence. The images share the same imaging path, camera, and light settings. For the current IMAST setting, the 0-2 group corresponds to 1900cycles/rad.

There are many ways to quantify image degradation. The most direct way, which also works well with SUIM, is to estimate the image degradation in terms of the MTF, which describes the total system response at different spatial frequencies. This might not be the best approach for turbulence-degraded images, especially under high levels of distortion. However, we intend to estimate the long-exposure (averaged) impacts, and such an approach is acceptable for this purpose. There are several methods to derive the MTF from imagery. We chose a standard slant-edge technique [20

20. I. A. Cunningham and A. Fenster, “A method for modulation transfer function determination from edge profiles with correction for finite-element differentiation,” Med. Phys. 14, (1987). [CrossRef]

,21

21. ISO, “Electronic still picture imaging spatial frequency response (SFR) measurements” (International Organisation for Standardization, 1997).

] by measuring the corresponding frame-averaged MTF. The detail of such an approach is not the focus of this paper. However, it is worth mentioning that efforts have been made to ensure error-free implementation by checking in air as well as particle- and turbulence-free underwater conditions.

To test our hypothesis that the differences in image quality shown at different depths (Fig. 5) is caused by the optical turbulence, we checked the single-frame MTF against multiframe-averaged MTF. Image degradation by particle scattering alone should remain the same between the single-frame and multiframe averaged MTFs, while the ones involving turbulence scattering degradation would be reduced in quality when multiple frames are averaged. This is due to the static nature of the PSF associated with the particle scattering process [16

16. W. Hou, D. Gray, A. Weidemann, and R. Arnone, “Comparison and validation of point spread models for imaging in natural waters,” Opt. Express 16, 9958–9965 (2008). [CrossRef]

] as a result of evenly distributed scattering centers—that is, particles. The larger turbulent cells, or turbules, and their temporal variations in size and shape, on top of uneven spatial distribution, determine the nonstatic nature of the PSF associated with optical turbulence. The results show that there is very little difference in image quality under the weak turbulence situation (2.8 m) but noticeable differences under the stronger turbulence case (8.7 m, Fig. 6), which confirms our hypothesis, at least to the first order.

Fig. 6. Normalized MTF of individual (light dashed) and 10-frame averaged (solid line) images obtained under strong (8.7 m) and weak (2.8 m) optical turbulence during SOTEX. MTFs are calculated using slant-edge algorithms over the same ROI for all images. The optical properties (particle scattering) of these images are similar and can be seen in Fig. 2.

From SUIM, we can see that if we assume the path radiance is the same, which is the main reason for conducting the night deployment with an active target, and we also assume the particulate scattering characteristics, including the single scattering albedo, ωo, and the mean scattering angle, are the same at the depths we are interested in, then the difference of the MTF at depth 2 (H2) relative to depth 1 (H1) can be written as
H2(ψ,r)=exp[(Sn2Sn1)ψ5/3r]exp[(c2c1)(1ϖ0(1e2πθ0ψ2πθ0ψ)]H1(ψ,r),
(3)
where Sn2 and Sn1 are the optical turbulence intensities at the corresponding depths, respectively.

From Eq. (3), we calculated MTFs for the depth of 8.7 m (H2) to include the impacts of optical turbulence, using measured optical properties and turbulence dissipation rates at this depth along with measured MTFs (H1) from the turbulence-free region (2.8 m). The results are shown in Figs. 7(a–d). The images used are obtained from an active source (one-way path) during the night deployment of July 27 to minimize path radiance. Figure 7(a) shows the averaged MTF at the shallower depth (2.8 m), where the optical turbulence is weak, compared to several cases (sequences A, B, and C as marked) at the deeper depth (8.7 m), where the optical turbulence is strong. For each of the image sequences at 8.7 m, the MTFs are estimated using the same region of interest (ROI) of consecutive frames, over 10 frames for long exposure. They are compared to the single-frame and averaged-frame results at 2.8 m, as well as the modeled outcome [Figs. 7(b), 7(c), and 7(d)]. As explained above, the averaging and comparison to single-frame results at 2.8 m is used to examine the level of degradation over longer exposure as a crude indicator of turbulence degradation (or lack thereof, in this case). The SUIM model [15

15. W. Hou, “A simple underwater imaging model,” Opt. Lett. 34, 2688–2690 (2009). [CrossRef]

,22

22. W. Hou, S. Woods, W. Goode, E. Jarosz, and A. Weidemann, “Impacts of optical turbulence on underwater imaging,” Proc. SPIE 8030, 883009 (2011). [CrossRef]

] is used to incorporate the impacts of optical turbulence at 8.7 m, using different turbulence parameters for each sequence, with R0=0.0045, 0.006, and 0.008m1. These values are calculated using Eq. (2), applying measured TKED and TD rates [23

23. S. Woods, W. Hou, W. Goode, E. Jarosz, and A. Weidemann, “Quantifying turbulence microstructure for improvement of underwater imaging,” Proc. SPIE 8030, 883009 (2011). [CrossRef]

] as ε=107, 109, 109.2m2s3 and χ=106, 106.4, 106.3°C2s1, respectively [19

19. S. Woods, Naval Research Laboratory, Stennis Space Center, MS 39529, USA, W. Hou, E. Jarosz, W. Goode, and A. Weidemann are preparing a manuscript be called “Measurements of turbulence dissipation in Lake Skeneateles.”

]. It is worth noting that the signal-to-noise ratio (SNR) cannot be improved when multiple frames are used, as each individual frame would typically undergo a different amount of degradation. Therefore the averaging would only increase the SNR towards the low-frequency elements and leave behind random variations at the high-frequency end. This is necessary, however, in order to contain all of the variations caused by the optical turbulence [15

15. W. Hou, “A simple underwater imaging model,” Opt. Lett. 34, 2688–2690 (2009). [CrossRef]

].

Fig. 7. MTFs of three different image sequences estimated from 8.7 m using the slant-edge method, and compared to the modeled results, during July 27 night deployment. (a) The relative variations in MTF of the three image sequences from 8.7 m depth (marked A, B, and C), compared to the averaged value at 2.8 m. (b–d) compare individual sequences at 8.7 m and 2.8 m respectively.

4. Discussion

One should bear in mind that even under turbid conditions, optical turbulence still affects image quality, although the relative contribution would be small. Although this is evident from Eq. (1), intuitively it is not necessarily the case, which is the primary reason behind this paper. By applying TKED rates in the range of 107 to 109m2s3, and temperature dissipation rates of 106 to 107°C2s1, one can see that the model result approximates the field measurements reasonably well. It is understandable that the model cannot be applied in all situations and should not be expected to fully explain every detail of the measurements, considering the challenges involved in quantifying turbulence with the IMAST [23

23. S. Woods, W. Hou, W. Goode, E. Jarosz, and A. Weidemann, “Quantifying turbulence microstructure for improvement of underwater imaging,” Proc. SPIE 8030, 883009 (2011). [CrossRef]

]. It would be beneficial, therefore, to examine discrepancies for the purpose of improvement in model closure as well as measurement accuracy.

The differences between the model and measurements could be the result of myriad factors. The primary reason is likely the difference in measurement scales in the temporal domain. In other words, the amount of time needed to measure the turbulence structure, both TKED and TD, is much longer than the time needed to observe the EO processes. This is understandable, as the only way to quantify chaotic processes is to use a statistical approach, which requires sufficient sampling points in time and space in order to eliminate impacts from random fluctuations. This is troublesome, especially when spatial fluctuations of the turbulent flow cannot be treated as isotropic nor statistically homogenous. This uncertainty caused by measurement mismatch will prevent a more in-depth understanding of the process involved, thus hindering efforts to mitigate the impacts. In theory, a short-exposure method can be used to freeze the turbulence impacts for restoration, similar to methods used in astronomy. However, it is very difficult to determine whether nearby distortions are from the same isoplanatic patch when relatively few data samples are available. Highly complex iterative methods involving the fusion of the optical flow method and a lucky-patch approach are under development in order to compensate for or reduce the degradation [24

24. A. Kanaev, W. Hou, and S. Woods, “Multi-frame underwater image restoration,” Proc. SPIE 8185, 818500 (2011). [CrossRef]

]. Because of this limitation, the best one can do for the model validation is probably using a through-the-sensor approach and range check as we have done here. In other words, due to the incompatibility in temporal scale measurements, we could not expect to pinpoint an exact value in TKED and TD rates but can only rely on the range of typical values. An instantaneous assessment method of the turbulence strength across the imaging path is highly desirable, which might stem from the success of the scintillation approach in astronomy and optical communications [25

25. G. Potvin, J. L. Forand, and D. Dion, “Some theoretical aspects of the turbulent point-spread function,” Appl. Opt. 24, 2932–2942 (2007). [CrossRef]

]. More rigorous postprocessing of images involving multiple ROIs, and direct derivation of MTFs using the whole resolution pattern, should help to improve the results as well. This would require high-resolution, high-speed imaging cameras. This is also the reason why we only presented data from one nightly deployment, as the spatial and temporal variability discussed above would override small environmental fluctuations between deployments.

Nevertheless, we can carry out a quick investigation by examining the degradation impacts and see if the above hypothesis carries any merits. We ran the SUIM with a different set of optical properties, thus different particulate scattering, to see if the model differences shown in Fig. 7 can be reduced. The results are shown in Fig. 8 for the three cases (A, B, and C) examined earlier. By introducing the amount of “excessive” scattering ranging from 0.16 to 0.2m1, a closer fit can be achieved [Figs. 8(a–c)], which supports our arguments above. Additionally, this also suggests that there might exist discrepancies between the time-invariant beam-c measurements and the time-varying degradation from the turbulence.

Fig. 8. (a)–(c) MTF model results improved by including contribution from extra particle scattering for each image sequence estimated from the depth of 8.7 m (marked A, B, and C)These are the same image sequences used in Fig. 7 (obtained during July 27 night deployment).

Our analysis of underwater images of beam propagation suggest that degradation by particle scattering and turbulence is likely a coupled process, especially under turbid conditions, where the movement of low index-of-refraction particles associated with turbulence kinetic-energy dissipation causes additional fluctuations in the image degradation. Efforts are underway to investigate and model such contributions. For the SOTEX study, the single scattering albedo at the depths examined remained relatively constant, around 0.75, which should not incur strong coupling effects on the image degradation.

5. Conclusion

The impact of optical turbulence on image degradation was directly measured in the field for the first time during SOTEX in July 2010, simultaneously with both optical properties and turbulence dissipation rates. This paper presented direct, quantified evidence of image degradation by varying strengths of optical turbulence in the natural environment, described in terms of MTFs, under similar turbidity conditions. By quantifying the degradation through image collection, field turbulent conditions through TKED and TD rates, and optical conditions through absorption and scattering, we achieved model closure by partitioning image degradation according to turbulence scattering and particle scattering, respectively. Results showed that directly measured MTFs over the turbulent flow using the slant-edge method from the field-collected images matched well with model-derived MTFs based on a combined model including impacts from both the particles and turbulence in the water. The discrepancies can be attributed to factors associated with differences in measurement scales in both temporal and spatial domains. Solutions to these issues are discussed. Further research is necessary to fully understand and address these topics. Efforts are also underway to investigate more complex turbulence conditions involving dissipation from both temperature and salinity variations in the ocean and to compensate for the degradation from optical turbulence on images.

This research was supported by ONR program element 62782N (NRL base project 73-6369). The authors thank the scientists and staff at the Upstate Freshwater Institute (UFI) for their assistance throughout SOTEX, and four reviewers for their insightful comments.

References

1.

G. D. Gilbert and J. C. Pernicka, “Improvement of underwater visibility by reduction of backscatter with a circular polarization technique,” Appl. Opt. 6, 741–746 (1967). [CrossRef]

2.

W. Hou, A. Weidemann, D. Gray, and G. R. Fournier, “Imagery-derived modulation transfer function and its applications for underwater imaging,” Proc. SPIE 6696, 6696221(2007). [CrossRef]

3.

W. Hou, Z. Lee, and A. Weidemann, “Why does the Secchi disk disappear? An imaging perspective,” Opt. Express 15, 2791–2802 (2007). [CrossRef]

4.

P. C. 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, 2794–2803 (2003). [CrossRef]

5.

L. Mullen, A. Laux, B. M. Concannon, E. P. Zege, I. L. Katsev, and A. S. Prikhach, “Amplitude-modulated laser imager,” Appl. Opt. 43, 3874–3892 (2004). [CrossRef]

6.

G. R. Fournier, D. Bonnier, J. L. Forand, and P. W. Pace, “Range-gated underwater laser imaging system,” Opt. Eng. 32, 2185–2190 (1993). [CrossRef]

7.

D. L. Fried, “Optical resolution through a randomly inhomogeneous medium for a very long and very short exposures,” J. Opt. Soc. Am. 56, 1372–1379 (1966). [CrossRef]

8.

M. C. Roggemann and B. M. Welsh, Imaging through Turbulence (CRC, 1996).

9.

N. Kopeika, “Imaging through the atmosphere for airborne reconnaissance,” Opt. Eng. 26, 1146–1154 (1987). [CrossRef]

10.

D. Sadot, A. Dvir, I. Bergel, and N. Kopeika, “Restoration of thermal images distorted by the atmosphere, based on measured and theoretical atmospheric modulation transfer function,” Opt. Eng. 33, 44–53 (1994). [CrossRef]

11.

Y. Yitzhaky, I. Dror, and N. Kopeika, “Restoration of atmospherically blurred images according to weather-predicted atmospheric modulation transfer functions,” Opt. Eng. 36, 3064–3072 (1997). [CrossRef]

12.

W. H. Wells, “Theory of small angle scattering,” in AGARD Lecture Series (NATO, 1973), Vol. 61.

13.

D. J. Bogucki, J. A. Domaradzki, R. E. Ecke, and C. R. Truman, “Light scattering on oceanic turbulence,” Appl. Opt. 43, 5662–5668 (2004). [CrossRef]

14.

G. D. Gilbert and R. C. Honey, “Optical turbulence in the sea,” in Underwater Photo-optical Instrumentation Applications (SPIE, 1972), pp. 49–55.

15.

W. Hou, “A simple underwater imaging model,” Opt. Lett. 34, 2688–2690 (2009). [CrossRef]

16.

W. Hou, D. Gray, A. Weidemann, and R. Arnone, “Comparison and validation of point spread models for imaging in natural waters,” Opt. Express 16, 9958–9965 (2008). [CrossRef]

17.

G. K. Batchelor, “Small-scale variation of convected quantities like temperature in turbulence fluid,” J. Fluid Mech. 5, 113–133 (1959). [CrossRef]

18.

S. W. Effler, A. R. Prestigiacomo, and D. M. O’Donnel, “Water quality and limnological monitoring for Skaneateles Lake: field year 2007” (Upstate Freshwater Institute, 2008), p. 57.

19.

S. Woods, Naval Research Laboratory, Stennis Space Center, MS 39529, USA, W. Hou, E. Jarosz, W. Goode, and A. Weidemann are preparing a manuscript be called “Measurements of turbulence dissipation in Lake Skeneateles.”

20.

I. A. Cunningham and A. Fenster, “A method for modulation transfer function determination from edge profiles with correction for finite-element differentiation,” Med. Phys. 14, (1987). [CrossRef]

21.

ISO, “Electronic still picture imaging spatial frequency response (SFR) measurements” (International Organisation for Standardization, 1997).

22.

W. Hou, S. Woods, W. Goode, E. Jarosz, and A. Weidemann, “Impacts of optical turbulence on underwater imaging,” Proc. SPIE 8030, 883009 (2011). [CrossRef]

23.

S. Woods, W. Hou, W. Goode, E. Jarosz, and A. Weidemann, “Quantifying turbulence microstructure for improvement of underwater imaging,” Proc. SPIE 8030, 883009 (2011). [CrossRef]

24.

A. Kanaev, W. Hou, and S. Woods, “Multi-frame underwater image restoration,” Proc. SPIE 8185, 818500 (2011). [CrossRef]

25.

G. Potvin, J. L. Forand, and D. Dion, “Some theoretical aspects of the turbulent point-spread function,” Appl. Opt. 24, 2932–2942 (2007). [CrossRef]

26.

W. Hou, “Characteristics of large particles and their effects on submarine light field,” Ph.D. dissertation (University of South Florida, 1997), p. 149.

OCIS Codes
(010.4450) Atmospheric and oceanic optics : Oceanic optics
(010.7060) Atmospheric and oceanic optics : Turbulence
(010.7295) Atmospheric and oceanic optics : Visibility and imaging

ToC Category:
Atmospheric and Oceanic Optics

History
Original Manuscript: October 12, 2011
Revised Manuscript: February 1, 2012
Manuscript Accepted: March 4, 2012
Published: May 10, 2012

Virtual Issues
Vol. 7, Iss. 7 Virtual Journal for Biomedical Optics

Citation
Weilin Hou, Sarah Woods, Ewa Jarosz, Wesley Goode, and Alan Weidemann, "Optical turbulence on underwater image degradation in natural environments," Appl. Opt. 51, 2678-2686 (2012)
http://www.opticsinfobase.org/vjbo/abstract.cfm?URI=ao-51-14-2678


Sort:  Author  |  Year  |  Journal  |  Reset  

References

  1. G. D. Gilbert and J. C. Pernicka, “Improvement of underwater visibility by reduction of backscatter with a circular polarization technique,” Appl. Opt. 6, 741–746 (1967). [CrossRef]
  2. W. Hou, A. Weidemann, D. Gray, and G. R. Fournier, “Imagery-derived modulation transfer function and its applications for underwater imaging,” Proc. SPIE 6696, 6696221(2007). [CrossRef]
  3. W. Hou, Z. Lee, and A. Weidemann, “Why does the Secchi disk disappear? An imaging perspective,” Opt. Express 15, 2791–2802 (2007). [CrossRef]
  4. P. C. 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, 2794–2803 (2003). [CrossRef]
  5. L. Mullen, A. Laux, B. M. Concannon, E. P. Zege, I. L. Katsev, and A. S. Prikhach, “Amplitude-modulated laser imager,” Appl. Opt. 43, 3874–3892 (2004). [CrossRef]
  6. G. R. Fournier, D. Bonnier, J. L. Forand, and P. W. Pace, “Range-gated underwater laser imaging system,” Opt. Eng. 32, 2185–2190 (1993). [CrossRef]
  7. D. L. Fried, “Optical resolution through a randomly inhomogeneous medium for a very long and very short exposures,” J. Opt. Soc. Am. 56, 1372–1379 (1966). [CrossRef]
  8. M. C. Roggemann and B. M. Welsh, Imaging through Turbulence (CRC, 1996).
  9. N. Kopeika, “Imaging through the atmosphere for airborne reconnaissance,” Opt. Eng. 26, 1146–1154 (1987). [CrossRef]
  10. D. Sadot, A. Dvir, I. Bergel, and N. Kopeika, “Restoration of thermal images distorted by the atmosphere, based on measured and theoretical atmospheric modulation transfer function,” Opt. Eng. 33, 44–53 (1994). [CrossRef]
  11. Y. Yitzhaky, I. Dror, and N. Kopeika, “Restoration of atmospherically blurred images according to weather-predicted atmospheric modulation transfer functions,” Opt. Eng. 36, 3064–3072 (1997). [CrossRef]
  12. W. H. Wells, “Theory of small angle scattering,” in AGARD Lecture Series (NATO, 1973), Vol. 61.
  13. D. J. Bogucki, J. A. Domaradzki, R. E. Ecke, and C. R. Truman, “Light scattering on oceanic turbulence,” Appl. Opt. 43, 5662–5668 (2004). [CrossRef]
  14. G. D. Gilbert and R. C. Honey, “Optical turbulence in the sea,” in Underwater Photo-optical Instrumentation Applications (SPIE, 1972), pp. 49–55.
  15. W. Hou, “A simple underwater imaging model,” Opt. Lett. 34, 2688–2690 (2009). [CrossRef]
  16. W. Hou, D. Gray, A. Weidemann, and R. Arnone, “Comparison and validation of point spread models for imaging in natural waters,” Opt. Express 16, 9958–9965 (2008). [CrossRef]
  17. G. K. Batchelor, “Small-scale variation of convected quantities like temperature in turbulence fluid,” J. Fluid Mech. 5, 113–133 (1959). [CrossRef]
  18. S. W. Effler, A. R. Prestigiacomo, and D. M. O’Donnel, “Water quality and limnological monitoring for Skaneateles Lake: field year 2007” (Upstate Freshwater Institute, 2008), p. 57.
  19. S. Woods, Naval Research Laboratory, Stennis Space Center, MS 39529, USA, W. Hou, E. Jarosz, W. Goode, and A. Weidemann are preparing a manuscript be called “Measurements of turbulence dissipation in Lake Skeneateles.”
  20. I. A. Cunningham and A. Fenster, “A method for modulation transfer function determination from edge profiles with correction for finite-element differentiation,” Med. Phys. 14, (1987). [CrossRef]
  21. ISO, “Electronic still picture imaging spatial frequency response (SFR) measurements” (International Organisation for Standardization, 1997).
  22. W. Hou, S. Woods, W. Goode, E. Jarosz, and A. Weidemann, “Impacts of optical turbulence on underwater imaging,” Proc. SPIE 8030, 883009 (2011). [CrossRef]
  23. S. Woods, W. Hou, W. Goode, E. Jarosz, and A. Weidemann, “Quantifying turbulence microstructure for improvement of underwater imaging,” Proc. SPIE 8030, 883009 (2011). [CrossRef]
  24. A. Kanaev, W. Hou, and S. Woods, “Multi-frame underwater image restoration,” Proc. SPIE 8185, 818500 (2011). [CrossRef]
  25. G. Potvin, J. L. Forand, and D. Dion, “Some theoretical aspects of the turbulent point-spread function,” Appl. Opt. 24, 2932–2942 (2007). [CrossRef]
  26. W. Hou, “Characteristics of large particles and their effects on submarine light field,” Ph.D. dissertation (University of South Florida, 1997), p. 149.

Cited By

Alert me when this paper is cited

OSA is able to provide readers links to articles that cite this paper by participating in CrossRef's Cited-By Linking service. CrossRef includes content from more than 3000 publishers and societies. In addition to listing OSA journal articles that cite this paper, citing articles from other participating publishers will also be listed.


« Previous Article  |  Next Article »

OSA is a member of CrossRef.

CrossCheck Deposited