Abstract
When imaging bright light sources, rays of light emanating from their centres are commonly observed; this ubiquitous phenomenon is known as the starburst effect. The prediction and characterization of starburst patterns formed by extended sources have been neglected to date. In the present study, we propose a novel trichromatic computational framework to calculate the image of a scene viewed through an imaging system with arbitrary focus and aperture geometry. Diffractive light transport, imaging sensor behaviour, and implicit image adjustments typical in modern imaging equipment are modelled. Characterization methods for key optical parameters of imaging systems are also examined. Extensive comparisons between theoretical and experimental results reveal excellent prediction quality for both focused and defocused systems.
Introduction
Captured images of light sources commonly exhibit the starburst effect, an optical phenomenon comprising apparent rays of light emanating from their centres. These rays, known as diffraction spikes, are also observable by the naked human eye, usually at night. Diffraction spikes in telescope images of stars and other illuminated bodies^{1,2,3} introduce uncertainties in luminositydependent measurements, but can be useful in localization techniques^{4}. The phenomenon occurs on all light sources and affects a wide range of imaging systems, including photography^{5,6,7}, medical endoscopy^{8}, and telemetry acquisition systems^{9}, with higherintensity sources yielding more prominent spikes.
Often accompanied with lens flare^{10,11}, the starburst effect arises due to the diffraction of light as it propagates past the limiting aperture of the imaging system^{12}. A Fourier optics formulation is typically employed, where the diffractionlimited point spread function is given by the Fourier transform of the exit pupil shape. It is common for imaging systems at high fnumbers to have polygonal apertures—these admit high spatial frequency components along axes perpendicular to the polygonal edges^{13,14}, hence forming the perceived spikes. In reflective telescopes, the support vanes of secondary mirrors result in a diffraction pattern similar to that formed by multiple intersecting slits^{15}.
The suppression and intensification of the starburst effect have received much attention to date. Efforts have been made to render visually similar effects in image postprocessing^{6} and minimize diffraction artifacts in high dynamic range (HDR) photography^{5}. In astronomy, software modelling Bahtinov masks and spiderdiffraction have been developed^{16}, and the reduction of diffractive effects on segmented mirrors is crucial for telescope design^{17}. Simulation toolsets and methods are also available for astronomical imagery, encompassing light transport effects including weak gravitational lensing and Doppler shift^{18,19,20}. Outside of astronomy, however, limited attention has been placed on correctly predicting the images of extended objects on general apertures and focus, with the optical parameters and implicit image processing of the imaging system taken into account; a lacuna in this discipline therefore remains. While predicting the corresponding image of a pointlike object entails a calculation of the pointspread function (PSF), predictions for extended objects require a convolution of the PSF with the object field, significantly increasing the complexity of the problem. Addressing this gap enables greater accuracy and generality in modelling the starburst effect, thereby enhancing its diverse applications, especially in image prediction on commercial imaging systems—such is the focus of our study.
This paper presents a rigorous framework for calculating the image of a scene viewed through an imaging system with arbitrary focus. A Fourier optics formulation is first discussed, followed by a computational image prediction framework. The characterization of a benchmark imaging system and the adopted experimental method are then described, with extensive comparisons between theoretical and experimental results.
Optics Formulation
In Fourier optics, a system of lenses and apertures can be reduced to corresponding entrance and exit pupils, wherein diffractive effects can be equivalently treated^{21}; the exit pupil is used throughout this paper. The media of the object and image spaces are taken to be identical, therefore the nodal points and principal planes are coincident^{22}. A plausible alternative to the adopted Fourieroptical formulation is Monte Carlo raytracing extended to model edge diffraction^{23}, though this remains outside the scope of the current study.
The system geometry is defined in Fig. 1. The geometrical coordinates on the image plane R are denoted (u, v). The heights d and d′ are defined at the entrance and exit pupils, as shown by the intersection of the limiting rays with reference spheres centered at the origins of the object and image planes respectively. The linear magnification of the system is then \(M=d{z}_{{\rm{i}}}/d^{\prime} {{z}}_{{\rm{o}}}\), where z_{o} and z_{i} are the distances between the object plane and entrance pupil, and between the image plane and the exit pupil, respectively.
Here an extended object at O is illuminated by spatially incoherent light. The actual image intensity distribution can be written \({I}_{i}(u,v)=h(u,v{)}^{2}\otimes {I}_{g}(u,v)\), where \(\otimes \) denotes a convolution, I_{g} is the magnified objectspace intensity distribution, and \(h{}^{2}\) and h are the intensity and amplitude point spread functions respectively. Note that h is complex, encoding both amplitude and phase information. To compute the convolution, the optical transfer function (OTF) \( {\mathcal H} = {\mathcal F} \{h{}^{2}\}\) given by
is utilized, where \(H= {\mathcal F} \{h\}\) is the amplitude transfer function (ATF) and \( {\mathcal F} \{\cdot \}\) denotes the twodimensional Fourier transform operator. Calculation of the OTF from a known ATF in this manner is known as the doubletransform method^{24}. The ATF of an imaging system takes the form
where \(k=2\pi /\lambda \) and P(x, y) is the exit pupil function describing the bounded pupil area \({{\mathscr{S}}}_{{\rm{p}}{\rm{u}}{\rm{p}}}\). The exponent in Equation (2) accounts for any phase shift kW(x, y) at the exit pupil due to aberrations. Considering onaxis image points (see Fig. A.1 of Supplementary Information), kW(x, y) can be obtained by subtracting the ideal phase distribution across the exit pupil from the actual one. Therefore,
where Δz is the distance between the infocus image plane F and the outoffocus image plane R. These calculations are valid in the paraxial approximation, an inherent limitation in Fourieroptical formulations.
For simple pupil shapes, the OTF can be solved analytically from Equation (1) with the normalized autocorrelation function of H ^{21,25}. For complex pupil shapes, either the doubletransform method or normalized autocorrelation may be performed numerically. While the former relies on fast Fourier transform (FFT) algorithms, the latter requires a polygon clipping algorithm^{26} for each sampling point of the OTF, to determine the area over which \(\exp [\iota kW(x,y)]\), a highly oscillatory function at large defocusing, may be numerically integrated. Such a procedure is both time and memory intensive; the doubletransform method is hence preferable.
The isoplanacity of the imaging lens system is assumed in this model as a simplification. A more complete treatment will entail the characterization of the modulation transfer function (MTF) of the lens system over the image area; such a method, however, yields a spatially variant PSF that is incompatible with a Fourieroptical formulation. To retain the low computational cost of Fourieroptical approaches, the imaging region of interest is taken to lie within the isoplanatic patch of the lens. This condition is expected to be satisfied if the paraxial condition holds.
Computational Model
In this section, the computational framework for the prediction of diffraction spikes is discussed. The process is divided into two primary segments (Fig. 2). First, the doubletransform method computes the theoretical channelspecific raw pixel values; a postprocessing pipeline then renders the final colour image. In this manner, the predicted colour image corresponds directly with images taken by imaging equipment, and the two can therefore be compared.
A trichromatic approach is adopted—predictions are made based on the composition of three monochromatic colour channels, each of a specific peak wavelength. This approximation yields a significant reduction in computational complexity, as compared to a polychromatic approach involving integration across the bandpasses of imaging sensor elements. In the adopted approach, the wavelengthdependent sensitivity of the sensor is simplistically treated via empirical characterization on each channel, as will be described in the next section; this implies that the full spectral power distribution of the source need not be known. A trichromatic approach may be analogous to the technological nature of modern imaging sensors and displays^{27}, and to biological vision in humans and many other organisms^{28,29,30}.
Raw Pixel Data Prediction
We seek to calculate the theoretical intensity distribution across the imaging sensor, such that the raw pixel values can be computed. First, the normalized object intensity \({I}_{o}\in \mathrm{[0,}\,\mathrm{1]}\) is represented by an m × n matrix A covering a geometrical size of \(2{\xi }_{{\rm{\max }}}\times 2{\eta }_{{\rm{\max }}}\). The geometrical image therefore has a size \(\mathrm{2}M{\xi }_{{\rm{\max }}}\times \mathrm{2}M{\eta }_{{\rm{\max }}}\) and a normalized intensity matrix B given by \({{\bf{B}}}_{i,j}={{\bf{A}}}_{isgnM,jsgnM}\). To calculate the OTF matrix \({\boldsymbol{ {\mathcal H} }}\), the pupil function matrix in the frequency domain \({\boldsymbol{\mathcal P}}\) is first constructed by scaling \({{\mathscr{S}}}_{{\rm{p}}{\rm{u}}{\rm{p}}}\) (defined in the spatial domain) by 1/λz_{i}, and uniformly sampling it m × n times within the domain of \({\boldsymbol{ {\mathcal B} }}= {\mathcal F} \{{\bf{B}}\}\). The ATF matrix is then
where \({f}_{u,{\rm{\max }}}=m\mathrm{/4}M{\xi }_{{\rm{\max }}}\) and \({f}_{v,{\rm{\max }}}=n\mathrm{/4}M{\eta }_{{\rm{\max }}}\) are the Nyquist frequencies along the u and v axes. The image must be of sufficiently high resolution such that the Nyquist frequency is larger than the cutoff frequency of the ATF. Sufficient null padding is also necessary for periodicity breaking. Utilizing the doubletransform method, the OTF matrix \({\boldsymbol{ {\mathcal H} }}\) can be computed as \({\boldsymbol{ {\mathcal H} }}= {\mathcal F} \{{ {\mathcal F} }^{1}\{{\bf{H}}{\}}^{2}\}\). The predicted image intensity distribution is then \({\bf{C}}={ {\mathcal F} }^{1}\{{\boldsymbol{ {\mathcal H} }}\,\circ \,{\boldsymbol{ {\mathcal B} }}\}\) where \(\circ \) denotes the Hadamard product of matrices.
The actual intensity distribution incident on the imaging sensor is therefore κC, where κ is a proportionality constant dependent on the system geometry. The raw pixel value matrix D is then
where t is the exposure time, \({\varphi }_{0}\) is a reference radiant exposure for normalization, and \({\mathscr{Z}}\) is the sensor response function. The form of \({\mathscr{Z}}\) is intrinsic upon the physical construction of the sensor, and \({\varphi }_{0}\) is specific for a given source and sensor configuration. The physical parameters which \({\mathscr{Z}}\) and \({\varphi }_{0}\) depend on are detailed in Supplementary Information B, alongside characterization methods in the next section.
Finally, an appropriate colour filter is applied to D, yielding the predicted channelspecific pixel values. The colour filter is dependent upon the construction of the imaging sensor, and can be represented by a Bayer pattern. This entire process is repeated for all three colour channels to yield a complete pixelwise prediction in colour filter array (CFA) form.
PostProcessing Rendering
Further stages of processing are required to apply various adjustments that are oftentimes implicit in modern imaging equipment. In the undemosaiced CFA form, white balancing is performed, followed by the execution of a demosaicing algorithm to yield a true colour RGB image^{31}. Colour space conversion is then applied for accurate display on a computer monitor or in print^{32}. Gamma correction may also be applied. The final result from this process is a predicted colour image that corresponds directly to one captured by a modern imaging system, computed from first principles—the excellent accuracy of this method is demonstrated in later sections.
Imaging System Characterization
Predicting the behaviour of an imaging system through the presented computational model requires several characteristic parameters of the system to be known. These include the linear magnification of the lens system and the positions and sizes of the entrance and exit pupils. The positions of the principal planes are also necessary to ascertain the lens geometry at different focusing distances. As a demonstration, a Nikon D7200 DSLR camera with an AF Nikkor 50 mm f/1.8D prime lens is used as a verification benchmark. However, it is worth noting that the model presented in this paper is applicable to arbitrary imaging systems in general.
Pupil & Principal Plane Locations
Characteristic data for the examined lens system is available in existing literature^{33}. The extracted pupil and principal plane positions relative to the sensor plane are presented in Table B.1 of the Supplementary Information. While these measurements are valid for the camera at infinity focus, changing the focusing distance of the lens will result in a shift of the pupil locations. This can be calculated by treating the compound camera lens as a single thin lens with the principal planes^{12}.
Focus Adjustment
The effective focal length (EFL) of the system is denoted f. Consider the distance between the front principal plane and the object plane on which the system is focused (hereinafter s, see Fig. 3), and the distance between the rear principal plane and the sensor plane (hereinafter \(s^{\prime} \)). Clearly \(s^{\prime} =f\) when the lens is focused at infinity; thus, from Table B.1, f = (51.5 ± 0.1) mm.
The examined system is equipped with a prime lens—that is, focusing is achieved by the simultaneous movement of all lens elements back and forth within the lens enclosure as a single system. To achieve a finite focusing distance r_{f} away from the sensor plane, the lens elements must shift forward by a distance \(\delta z\) (Fig. 3). Denote the distance between P and P′ as z_{p}. Then, \(s={r}_{{\rm{f}}}{z}_{{\rm{p}}}s^{\prime} \) and \(s^{\prime} =f+\delta z\). Therefore,
from which δz and Δz as defined in Fig. A.1 can be calculated for pathlength error computation (see Supplementary Information A). This characterization method can be applied to prime lens systems in general. Zoom lenses are more complex due to their adjustable focal lengths; more detailed specifications are necessary for a complete characterization.
Pupil Sizes
The pupil sizes d and d′ remain to be determined. The geometry of Fig. 1 indicates \(d=\varepsilon {z}_{{\rm{o}}}/({\varepsilon }^{2}+{z}_{{\rm{o}}}^{2}{)}^{\mathrm{1/2}}\) and \(d^{\prime} =\varepsilon ^{\prime} {z}_{{\rm{i}}}/({\varepsilon ^{\prime} }^{2}+{z}_{{\rm{i}}}^{2}{)}^{\mathrm{1/2}}\), where ε and ε′ are the reference length scales measured from the optical axis to the farthest point on the pupil. To measure ε and ε′, the lens was detached from the camera body, and a second camera was used to photograph the entrance and exit pupils. The shapes and sizes of the pupils were determined using computational edgedetection (see Fig. B.1 of the Supplementary Information), on which \({{\mathscr{S}}}_{{\rm{pup}}}\), ε and ε′ can be defined.
Sensor Response
For incident light of intensity I_{0} and wavelength λ_{0}, dimensionless relative radiant exposures can be mapped to pixel values \(\zeta \in \mathrm{[0,}\,\mathrm{1]}\) via the sensor response function \({\mathscr{Z}}\) as follows:
where t is the exposure time. The physical validity of this mapping is detailed in Supplementary Information B. The trichromatic approach adopted therefore requires only empirical determination of \({\mathscr{Z}}\) and \({\varphi }_{0}\) for each colour channel. Figure 4 shows the individual response curves for each colour channel obtained by capturing images of the object at different exposure times and averaging the channelspecific pixel value within a fixed predefined region on the image. The nonlinearity at low exposure times is attributed to semiconductor nonidealities. The peak wavelengths of each channel (600 nm, 530 nm, and 450 nm for R, G and B respectively) were estimated based on existing quantum efficiency curves of similar camera models^{34}.
Model Verification
To demonstrate the accuracy of the presented computational method, sets of theoretical predictions are compared against experimental measurements. The benchmark experiment setup is first described, followed by comparisons between experimental results and theoretical predictions.
Experimental Setup
The characterization of the imaging device used has previously been detailed. The object in the benchmark experiment comprised a 20W LED floodlight with 5000 K colour temperature, over which a diffuser film was placed. A circular mask of diameter 10.5 mm was then mounted to create a uniform circular light source. The distance from the object to the imaging sensor plane was constant at 100.0 ± 0.1 cm. All experiments were conducted in a dark room to minimize ambient light pollution of images. The object occupied approximately 10% of the captured image area, and was axially aligned to ensure it remained within the isoplanatic patch.
The postprocessing applied to these experimental images and theoretical predictions were identical. In particular, an RGGB Bayer pattern was used, and a white balance of [R, G, B] = [2, 1, 2] was applied to account for the ratio of red, green and bluechannel pixels on the imaging sensor. Conversion to geometrical scale was performed using a pixel size of 3.917 × 3.917 μm^{35}. A gamma compression was also applied in accordance with the sRGB standard (see Supplementary Information A). No additional scaling was performed; the accuracy of the calculated magnification can thus be verified.
Results
Corresponding theoretical predictions were computed using the presented method. Matrix sizes of m = n = 3750 were adopted for sufficient null padding, followed by a cropping to 1250 × 1250 px to enhance visual clarity.
We compare theoretical predictions and experimental results for both a focused, diffractionlimited image as well as a severely defocused image. For each set of comparisons, image intensity profiles are plotted with circular as well as crosssectional sampling (Fig. 5). Comparisons between rendered sRGB colour images are also presented (Fig. 6), for which the Mean Square Error (MSE)^{36} and Structural Similarity Index (SSIM)^{37} are used as quantitative measures for the accuracy of the predictions.
Focused Images
The focused configuration comprises r_{o} = r_{f} = 100.0 cm, t = 2.5 s, and a f/22 aperture. Excellent agreement on image intensity profiles can be observed in Fig. 5(a–c). Clearly noticeable in Fig. 5(a) is the number of distinct diffraction spikes (14), twice that of the number of edges on the polygonal aperture (7). Indeed, for evensided apertures, the number of observable diffraction spikes will typically be identical to the number of aperture edges; whereas for oddsided apertures, the number of diffraction spikes will be twice the aperture edge count. A qualitative explanation is presented in Supplementary Information A.
Defocused Images
The defocused configuration comprises r_{o} = 100.0 cm, r_{f} = 45.0 cm, t = 1.6 s, and a f/22 aperture, corresponding to a 4.78 mm defocus. Excellent accuracy of theoretical predictions can be observed in Fig. 5(d–f). A comparison between Fig. 5(a) and 5(d) makes explicit the effects of defocus on the starburst effect, where for similar sampling radii, the inner “spikes” are blurred out and merge with the larger “rays” emanating from the centre of the imaged light source. This effect is well visualized when comparing the rendered focused and defocused images in Fig. 6. Clearly, a good overall agreement has been achieved.
Colour Images
Figure 6 compares the final predicted colour images from the computational method with experiments, for both the focused and defocused imaging system configuration. A slight alternation in the intensity of the diffraction spikes is visible in Fig. 6(b), suggesting nonidealities in the MTF of the lens system used. Excellent colour agreement is evident for both focused and defocused images. In addition, excellent results are obtained for the MSE (<0.1%) and SSIM (>95%) comparison tests, detailed in Table 1.
Conclusion
This paper has discussed a rigorous framework for calculating the image of a starburstaffected scene viewed through an imaging system with arbitrary focus and aperture geometry, based fundamentally on a physicallyvalid Fourier optics formulation. Following the computation of channelspecific pixel values via a numerical doubletransform method, we also propose a postprocessing pipeline accommodating various image adjustments standard in modern imaging equipment. The final result is a direct analogue of images captured by imaging apparatus, inclusive of light transport effects within the apparatus, demosiacing mechanisms on the imaging sensor, and implicit programmatic image adjustments. Such comprehensiveness represents a key advancement over existing literature.
Notably, excellent agreement between predictions and realworld experimental measurements were observed, for both focused and defocused configurations. These benchmark results indicate outstanding accuracy of our computational method. The trichromatic approach enables the prediction of blurred colour images unachievable with existing monochromatic, unaberrated methods; the required characterization of light sources with unknown spectral power distributions is also greatly simplified, a notable advantage over full polychromatic approaches. Our study has also presented key methods for the characterization of optical parameters of imaging systems, including edgedetection techniques for pupil geometry determination and the measurement of sensor response curves—these methods can be utilized to characterize other systems, on which computational predictions can then be made.
The presented results are of great relevance to the modelling and reduction of diffraction spikes in telescope and telemetry acquisition systems, critical for data accuracy^{1,2,3,9}; the Bahtinov maskfacilitated automated focusing of telescopes may also be improved, and spiderdiffraction phenomena typically encountered in reflecting telescopes may also be optimized^{16}. The framework may also be applied for the accurate generation of starburst visual effects in photorealistic computer graphics and the computation of diffraction spikes observed by the human eye, taking into account aberration effects^{38,39}. Extension of the computational framework to model xray diffraction^{40,41,42} and complex diffraction phenomena^{43,44,45} is also plausible, due to its mathematical and structural generality.
Data Availability
The datasets are available from the corresponding author on reasonable request.
References
 1.
Burrows, C. J. et al. The imaging performance of the hubble space telescope. The Astrophysical Journal 369, L21–L25 (1991).
 2.
Doi, K., Kaji, A., Takizawa, T. & Sayanagi, K. The application of optical transfer function in radiography. Japanese Journal of Applied Physics 3, 183 (1964).
 3.
Grady, C. A. et al. Hubble space telescope space telescope imaging spectrograph coronagraphic imaging of the herbig ae star ab aurigae. The Astrophysical Journal Letters 523, L151 (1999).
 4.
Heap, S. R. et al. Space telescope imaging spectrograph coronagraphic observations of β pictoris. The Astrophysical Journal 539, 435 (2000).
 5.
Lucat, A., Hegedus, R. & Pacanowski, R. Diffraction effects detection for hdr imagebased measurements. Optics express 25, 27146–27164 (2017).
 6.
Liu, D., Geng, H., Liu, T. & Klette, R. Stareffect simulation for photography. Computers & Graphics 61, 19–28 (2016).
 7.
Luksch, C. Realtime hdr rendering. Graduation Project. Institute of Computer Graphics and Algorithms, TU Vienna (2007).
 8.
Cameron, J. R., Skofronick, J. G. &Grant, R. M. Medical Physics: physics of the body. (Medical Physics Publishing Corporation, 1992).
 9.
Stiltz, H. L. Aerospace telemetry, vol. 1 (Prentice Hall, 1961).
 10.
Hullin, M., Eisemann, E., Seidel, H.P. & Lee, S. Physicallybased realtime lens flare rendering. ACM Trans. Graph. 30, 108:1–108:10 (2011).
 11.
Xiao, F., DiCarlo, J. M., Catrysse, P. B. & Wandell, B. A. High dynamic range imaging of natural scenes. 2002, 337–342 (2002).
 12.
Pedrotti, F. L. & Pedrotti, L. S. Introduction to optics 2nd edition. Introduction to Optics 2nd Edition by Frank L. Pedrotti, SJ, Leno S. Pedrotti New Jersey: Prentice Hall, 1993 (1993).
 13.
Huang, J. G., Christian, J. M. &McDonald, G. S. Fresnel diffraction and fractal patterns from polygonal apertures. J. Opt. Soc. Am. A 23, 2768–2774 (2006).
 14.
Lee, S.W. & Mittra, R. Fourier transform of a polygonal shape function and its application in electromagnetics. IEEE Transactions on Antennas and Propagation 31, 99–103 (1983).
 15.
Born, M. & Wolf, E. Principles of optics, cambridge university press. Cambridge, UK (1999).
 16.
Harvey, J. E. & Ftaclas, C. Diffraction effects of telescope secondary mirror spiders on various imagequality criteria. Appl. Opt. 34, 6337–6349 (1995).
 17.
Harvey, J. E., Irvin, R. G., Crabtree, K., Pfisterer, R. N. & Breckinridge, J. B. Diffraction analysis of large segmented mirror concepts for exoplanet exploration. Proc. SPIE 10698, 12, https://doi.org/10.1117/12.2313999 (2018).
 18.
Rowe, B. T. P. et al. Galsim: The modular galaxy image simulation toolkit. Astronomy and Computing 10, 121–150, http://www.sciencedirect.com/science/article/pii/S221313371500013X (2015).
 19.
Bridle, S. et al. Results of the great08 challenge*: an image analysis competition for cosmological lensing. Monthly Notices of the Royal Astronomical Society 405, 2044–2061 (2010).
 20.
Becker, M. R. calclens: weak lensing simulations for largearea sky surveys and secondorder effects in cosmic shear power spectra. Monthly Notices of the Royal Astronomical Society 435, 115–132 (2013).
 21.
Goodman, J. W. Introduction to Fourier optics (Roberts and Company Publishers, 2005).
 22.
Hecht, E. Hecht optics. Addison Wesley 997, 213–214 (1998).
 23.
Freniere, E. R., Gregory, G. G. & Hassler, R. A. Edge diffraction in monte carlo ray tracing. Proc. SPIE 3780, 7, https://doi.org/10.1117/12.363773 (1999).
 24.
Williams, C. S. & Becklund, O. A. Introduction to the optical transfer function. (Wiley New York etc, 1989).
 25.
Stokseth, P. A. Properties of a defocused optical system. J. Opt. Soc. Am. A 59, 1314–1321 (1969).
 26.
Maillot, P.G. A new, fast method for 2d polygon clipping: Analysis and software implementation. ACM Trans. Graph. 11, 276–290 (1992).
 27.
Sharma, G. & Trussell, H. J. Digital color imaging. IEEE Transactions on Image Processing 6, 901–932 (1997).
 28.
Jacobs, G. H. Evolution of colour vision in mammals. Philosophical Transactions of the Royal Society B: Biological Sciences 364, 2957–2967 (2009).
 29.
Surridge, A. K., Osorio, D. & Mundy, N. I. Evolution and selection of trichromatic vision in primates. Trends in Ecology & Evolution 18, 198–205 (2003).
 30.
Solomon, S. G. & Lennie, P. The machinery of colour vision. Nature Reviews Neuroscience 8, 276 (2007).
 31.
Malvar, H. S., Wei He, L. & Cutler, R. Highquality linear interpolation for demosaicing of bayerpatterned color images. In 2004 IEEE International Conference on Acoustics, Speech, and Signal Processing, vol. 3, iii4858 vol.3 (2004).
 32.
Sumner, R. Processing RAW Images in MATLAB (2014). Color space conversion algorithm retrieved from RAW Processing Guide, http://www.rcsumner.net/raw_guide/RAWguide.pdf, accessed (2018).
 33.
Wisniewski, J. S. Lens measurement table (2003). Data retrieved from Photography Resources, http://www.swissarmyfork.com/lens_table_1.htm, accessed (2012).
 34.
Darrodi, M. M., Finlayson, G., Goodman, T. & Mackiewicz, M. Reference data set for camera spectral sensitivity estimation. J. Opt. Soc. Am. A 32, 381–391 (2015).
 35.
Nikon Corporation. D7200 Digital SLR Camera (2015). Data retrieved from Specifications, http://imaging.nikon.com/lineup/dslr/d7200/spec.htm, accessed (2018).
 36.
Eskicioglu, A. M. & Fisher, P. S. Image quality measures and their performance. IEEE Transactions on Communications 43, 2959–2965 (1995).
 37.
Wang, Z., Bovik, A. C., Sheikh, H. R. & Simoncelli, E. P. Image quality assessment: from error visibility to structural similarity. IEEE Transactions on Image Processing 13, 600–612 (2004).
 38.
Liao, M. et al. Training to improve contrast sensitivity in amblyopia: correction of highorder aberrations. Scientific reports 6, 35702 (2016).
 39.
Liu, R. et al. Immature visual neural system in children reflected by contrast sensitivity with adaptive optics correction. Scientific reports 4, 4687 (2014).
 40.
Miao, J., Charalambous, P., Kirz, J. & Sayre, D. Extending the methodology of xray crystallography to allow imaging of micrometresized noncrystalline specimens. Nature 400, 342 (1999).
 41.
Chapman, H. N. & Nugent, K. A. Coherent lensless xray imaging. Nature Photonics 4, 833 (2010).
 42.
Tegze, M. & Faigel, G. Xray holography with atomic resolution. Nature 380, 49 (1996).
 43.
GonzalezIzquierdo, B. et al. Towards optical polarization control of laserdriven proton acceleration in foils undergoing relativistic transparency. Nature Communications 7, 12891 (2016).
 44.
Wu, K. & Wang, G. P. Twodimensional fibonacci grating for farfield superresolution imaging. Scientific reports 6, 38651 (2016).
 45.
Wu, K., Cheng, Q., Shi, Y., Wang, H. & Wang, G. P. Hiding scattering layers for noninvasive imaging of hidden objects. Scientific reports 5, 8375 (2015).
Author information
Affiliations
Contributions
M.L. performed theoretical and experimental work, wrote the computation code, analyzed the experimental and theoretical data, and wrote the paper. J.S.Q.T. performed theoretical and experimental work, wrote the computation code, analyzed the experimental and theoretical data, and wrote the paper. J.M.K. performed theoretical and experimental work, wrote the computation code, and wrote the paper. K.H.C. analyzed the results, wrote the paper, and directed the project.
Corresponding author
Ethics declarations
Competing Interests
The authors declare no competing interests.
Additional information
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Electronic supplementary material
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Lendermann, M., Tan, J.S.Q., Koh, J.M. et al. Computational Imaging Prediction of StarburstEffect Diffraction Spikes. Sci Rep 8, 16919 (2018). https://doi.org/10.1038/s4159801834400z
Received:
Accepted:
Published:
Keywords
 Diffraction Spikes
 Starburst Events
 Modern Imaging Equipment
 Arbitrary Focus
 Aperture Geometry
Further reading

What we have learnt from 30 years living with positive dysphotopsia after intraocular lens implantation?: a review
Expert Review of Ophthalmology (2021)

Comprehensive modelling and optical analysis of singlelens telescopes
Chinese Journal of Physics (2019)
Comments
By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.