JWST High-Contrast Imaging Roadmap
A roadmap to guide users, step-by-step, through the process of designing a JWST high-contrast imaging (HCI) observing program.
See also: Getting Started with JWST Proposing
High-contrast imaging (HCI) observations can be some of the most complex to schedule with JWST and for that reason the workflow of this roadmap is considered iterative. When planning HCI observations many parameters come into play and for some science cases, it is not always initially apparent which HCI mode—if at all—will provide you with the best scientific results; users may find themselves returning to earlier steps and/or stages before "linearly" producing their proposal and Astronomer's Proposal Tool (APT) files.
The roadmap points below are supplemented with links to articles for additional information.
Stage 1 – Become familiar with HCI capabilities and instrument-specific modes
In addition to the steps suggested in the article Getting Started with JWST Proposing (see the section "Become familiar with JWST capabilities, terminology, and documentation"), users should consider the following issues for high-contrast Imaging.
1. Which JWST observing modes enable HCI?
- MIRI coronagraphic imaging between 10 and 23 µm.
- NIRCam coronagraphic imaging between 1.8 and 5 µm.
- NIRISS aperture masking interferometry enabling high spatial resolution, moderate-contrast imaging between 2.7 and 4.8 µm.
- Imaging & IFU spectroscopy with non-coronagraphic PSF subtraction strategies enabling moderate-contrast imaging*.
2. What HCI optical designs are offered by JWST?
Familiarize yourself with the advantages, limitations, and functionality of each HCI design, as well as which scientific investigations they are optimized for.
- Lyot-type coronagraph: 5 implementations in NIRCam, one in MIRI.
- Four-quadrant phase-mask coronagraph (4QPMC): 3 implementations in MIRI.
- Non-redundant mask (NRM): one implementation in NIRISS.
3. What are the allowed mask-filter combinations for each of the HCI modes?
- MIRI, focal plane coronagraph masks: 3 4QPMs and one Lyot spot, each of which images to a different pupil mask and coronagraphic filter combination.
- NIRCam, sets of 5x occulting masks: 3 round and 2 bar-shaped, usable with a subset of permitted filters that depend on mask selection.
- NIRISS, a 7 hexagonal hole: (generating 21 baselines) non-redundant mask in the pupil plane usable with 4 NIRISS filters to enable AMI mode.
4. What are the primary performance metrics for HCI?
5. What are the predicted performances† of the instrument-specific modes?
- MIRI, achievable IWAs of 0.34–2.16": (1 λ/D for 4QPMs, 3 λ/D for Lyot) typical contrasts‡ achieve 10-4 to 10-5 for separations larger than 0.5″–1″.
- NIRCam, achievable IWAs of 0.14–0.89": (round and bar-shaped occulters optimized for 6 λ/D and 4 λ/D, respectively) contrasts typically ∼10−6 or better at 1′′ IWA and beyond.
- NIRISS AMI, achievable IWAs of 0.089–0.15": (1 λ/D) typical contrasts ~10-4 at separations of ~70–400 mas.
- Imaging and IFU spectroscopy: (with non-coronagraphic PSF subtraction) achievable contrasts ~10-3 to 10-4 for IWAs somewhere between those of AMI and coronagraphy for a given filter.
6. What are the fundamental physical limits for detection?
- Photon noise of the stellar point spread function (PSF):
- Detector noise:
- Background noise (zodiacal light + thermal emission), especially longward of ~15 μm.
7. What are the operations unique to HCI?
- MIRI coronagraphic imaging: target acquisition, HCI small grid dithers
- NIRCam coronagraphic imaging: target acquisition, coronagraphic PSF estimation, astrometric confirmation images, small grid dither
- NIRISS AMI: target acquisition, dithers (not recommended), and observing calibrators close in time to the targets.
8. What are the recommended observing strategies pertaining to HCI?
* HCI can be carried out using basic imaging modes of the observatory (Rajan et al., 2015; Durcan, Janson, and Carson, 2016), as well as using IFU strategies similar to Konopacky et al. (2013), however these modes are not yet covered in the documentation.
† Based on performance simulations and contrast predictions based on the latest information on the as-built telescope and instrument properties, including both static and dynamic contributions to wavefront error (Perrin et al. 2018)
‡ We report all contrasts as 5σ post-processing contrasts after single reference star subtraction.
Stage 2 – Compare your parameter space to the performance limits and capabilities of HCI observing modes
1. Identify wavelength range(s) of interest for your intended science.
How does this influence (or limit) your choice of science instrument(s), mask(s) and filter(s)?
- MIRI coronagraphic imaging: 3 4QPMs operating with narrowband filters centered at 10.65, 11.4, and 15.5 μm and one Lyot coronagraph working in a broadband filter centered at 23 μm.
- NIRCam coronagraphic imaging: one extra-wide, 4 wide-, 10 medium- and 2 narrowband filters (depending on the selection of coronagraphic mask), in the wavelength range 1.82–5.0 µm.
- NIRISS AMI: in order to enable AMI mode, the NRM will be used in conjunction with one of the 3 medium-band filters centered at 3.8, 4.3 and 4.8 μm (F380M, F430M, F480M) or a wideband filter centered at 2.77 μm (F277W).
2. Determine the apparent separations between host and companion source(s) at the time of observation.
Which instrument(s) and mask-filter combination(s) can achieve the required working angles?
- HCI Inner Working Angle article
- Throughput vs. apparent separation for combinations of coronagraphic mask and filter
3. Determine the companion contrast(s) at the wavelength(s) of interest.
Are your observations feasible given the contrast limits of the instrument(s)?
Note: when referring to a companion, the term "contrast" corresponds to the ratio of the companion's observed flux to that of its host. An observation is estimated to be feasible if the companion-to-host flux ratio is greater than the "limiting contrast" Climit(s).
Modeling may be required to extrapolate shorter wavelength measurements in the 3–23 μm regime for these predictions (e.g., to determine companion contrasts at MIRI wavelengths from far infrared or submillimeter data).
4. For coronagraphic observations, how important is the azimuthal coverage around your science target?
MIRI coronagraphic imaging:
- with the 4QPMs, the linear boundaries between adjacent quadrants attenuate light, reducing sensitivity in the field along the 4 edges of the mask.
- with the Lyot coronagraph, the Lyot spot is suspended in the focal plane by 2 supporting struts in the mounting bracket, which themselves block light in the FOV.
NIRCam coronagraphic imaging:
- the round occulting masks provide 360o azimuthal coverage around the bright object.
- the bar occulting masks sacrifice some FOV in the direction along the bar, as a function of azimuth around the bright object.
5. Is it possible that your scientific goals can be achieved with non-coronagraphic PSF subtraction?
For moderate contrasts (~ 10–3 to 10–4) and/or point source detections well in the background limited regime, it might be wise to opt for one of the standard imaging modes.
Stage 3 – Select a PSF calibration strategy
All HCI observations with JWST require the measurement and calibration of stellar point spread functions (PSFs) in some way for post-processing contrast reduction. For any PSF calibration strategy, the observing and data processing techniques are interdependent.
Coronagraphic PSF subtraction strategies
In order to achieve the necessary high contrast and recover faint sources surrounding the science target, one must calibrate and subtract the PSF of the central source
1. Consider the degrading factors that may limit the PSF calibration and what steps you will take to mitigate them.
These include wavefront drifts of the observatory, PSF star color differences, self-subtraction biases (especially for disks), imperfect target acquisitions, line-of-sight jitter, and dynamic wavefront error.
2. Which observing technique(s) will you include in your PSF subtraction strategy?
Each PSF calibration and subtraction method has a corresponding observing strategy. The imaging techniques are combined with post-processing optimization algorithms (such as LOCI§ and KLIP¶) to generate an optimal synthetic reference PSF, to be subtracted from the science target image.
Will you employ the "referenced differential imaging" (RDI) technique? (Required)
In this techniquethe observation of a nearby star is used to generate an unresolved, high signal-to-noise (SNR) PSF to subtract from the science target.
The RDI technique is sensitive to wavefront drifts and PSF star color differences.
By scheduling the science and PSF reference observations back-to-back in a sequence, the effect of wavefront drifts should be minimized.
Will you employ the "angular differential imaging" (ADI) technique? (Recommended)
In this technique, the science target is observed at 2 different roll angles and is used as a self-reference for PSF subtraction.
ADI allows for PSF subtraction at nearly the same spacecraft attitude (for wavefront stability) and helps mitigate detector artifacts.
However ADI comes at the cost of self-subtraction biases, especially given the limited available roll (~10°) of JWST.
For robustness, observations using both RDI and ADI PSF calibration techniques are strongly recommended. You can deviate from this plan if desired, but you must explain your alternate PSF subtraction strategy in your proposal.
More information here: standard coronagraphic sequence.
Will you employ the "small grid dithering" (SGD) technique? (Optional)
This technique involves p
The SGD technique is optional, and should only be used when the highest quality PSF subtraction is needed.
Non-coronagraphic PSF subtraction
Using the same PSF subtraction methods, it is also possible to achieve high performance with non-coronagraphic imaging modes, such as direct imaging in filters that may not have coronagraphs available, or using one of JWST’s integral field spectrographs in NIRSpec or MIRI. The contrasts achieved with such modes, even with careful PSF calibration, will not equal the contrasts achieved with the coronagraphs, but even “moderate” contrasts can still offer compelling science capabilities. Such observations are already planned for Cycle 1 by both GTO and ERS teams.
Interferometry calibration strategies
Which observing technique(s) will you include in your PSF calibration strategy?
Will you observe a PSF reference calibrator?
calibrator star allows for the instrumental systematics (affecting interferometric observables) to be measured. The reference PSF is used to calibrate out the instrumental contributions to closure phases (CP; the sum of three phases around a closed triangle of baselines) and squared visibility amplitudes (SqV), during the third stage in the calibration pipeline (CALWEBB_AMI3).
Near-contemporaneous acquisition of target and point source calibrator data is desirable, except for very low contrast needs: if contrast limits are not very demanding, a reference star from an unrelated observation, or possibly an analytically generated reference PSF can be used.
Will you employ the Kernel phase imaging technique?
§ “locally optimized combination of images” or LOCI algorithm (Lafrenière et al. 2007)
¶ KL image projection (KLIP) algorithm (Soummer et al. 2012)
Stage 4 – Assess target visibilities and allowed position angles
The following steps should be used in conjunction with those outlined in the Getting Started with JWST Proposing (see the section, "Determine if your targets can be observed").
1. Familiarize yourself with JWST position angles, coordinate systems, and related nomenclature to understand the telescope’s pointing constraints.
2. Determine the viewing constraints placed on your target(s).
3. Using at least one of the JWST target visibility tools, assess your target visibilities, as well as allowed position angles versus time.
4. In the case of known or expected companions, consider whether your observations require any restrictions on the orientation of the instrument field of view (FOV)/ detector being referenced.
This is especially in relation to any instrumental obstructions, such as the cross pattern for the MIRI FQPM, bars for NIRCam, or outside of gaps in the uv-pane coverage for NIRISS.
For coronagraphy, the CVT provides visualizations of the focal plane projected onto the sky, which is useful for evaluating the placement and orientation of known science sources on the coronagraphic masks.
Determine the aperture position angle(s)/ date(s) at which your companion(s) are nominally visible.
5. If implementing the ADI technique in your PSF calibration strategy (during the Select a PSF calibration strategy stage):
Check how the instantaneous roll flexibility changes over the the particular visibility period.
This instantaneous roll flexibility is approximately ±5° from nominal, but varies with time and look direction between ±3.5° to ±7°.
Assess how potential rolls of the telescope will change the positions of the companion(s) relative to any instrumental obstructions.
6. In the case of coronagraphy, consider whether your goals call for a larger roll offset on the science target than can be obtained instantaneously in a single visibility period.
Certain science cases may require a follow-up at some more substantial angular offset (e.g., 30° offset) relative to the first observations—for instance, to recover part of the scene that may be blocked by the selected mask. In such cases, the observation will have to be scheduled at a significantly later time.
Coronagraphic observers will want to assess their potential targets carefully, and when possible, select targets above 45° ecliptic latitude if they require large offsets in PA between observations.
The Coronagraphic Visibility Tool (CVT) can be used to assess the availability of multiple position angles, and estimate what the time separation will be. It is expected that each sequence will contain an observation of a relevant PSF reference star, since the PSF will likely change between the two epochs.
Stage 5 – Use the Exposure Time Calculator to determine observing parameters
Estimating exposure times is a science-critical aspect of HCI observation planning. Once target visibility is confirmed and a PSF calibration strategy adopted, the JWST Exposure Time Calculator (ETC) should be used to determine the exposure parameters needed to achieve the desired signal to noise (SNR) on your target(s). Aside from the directions in the Getting Started with JWST Proposing article, the following are advisable for HCI:
1. Define your scenes and sources.
Create a science scene in the ETC and populate it with the source targets.
The science scene should contain the source targets for observation and all other nearby sources that could contribute to both the observed target and background fluxes. The bright ("host") source should be placed in the center of the scene and the reference PSF source(s) at a significant offset (e.g., 10").
Faint sources must be placed within a square centered on the scene center for each instrument/mode pairing.
- for MIRI the scene is a 8.91" square centered on the coronagraphic masks.
- for NIRCam the scene is a 6.36" square centered on the coronagraphic masks.
- for NIRISS the scene is a 5.31" square centered on the NRM.
Create a reference scene and populate it with the reference PSF source.
In order to facilitate target acquisition calculations for the reference PSF source, a dedicated "reference" scene—containing only the reference PSF target—is required. The reference PSF source should be positioned at the center of the scene (offset 0,0).
2. Initiate calculations for each of your planned observations.
Words in bold italics are buttons
or parameters in GUI tools. Bold
style represents GUI menus/
panels & data software packages.
3. Adjust the exposure time via the NUMBER OF GROUPS, INTEGRATIONS, and/or EXPOSURES until you obtain the desired SNR and contrast on your target.
For NIRISS AMI, there is an advanced Python tool and precomputed contrast curves found at https://github.com/agreenbaum/ami_sim and hbp://maestria.astro.umontreal.ca/niriss/AMIcontrast/index.php, respectively.
4. Check your individual calculations for detector saturation.
While some saturation may be tolerable, only partially-saturated pixels will be recoverable. In saturated regions, the photometric accuracy will be suboptimal and the contrast will most likely be affected at or close to the IWA; consequently, faint portions of the astronomical scene that overlap with saturated pixels may not be properly detected.
Using the Saturation Map, check if any saturated pixels overlap with faint sources/features in the astronomical scene.
In the event of saturation, it may be possible to recover pixels at the expected position of the companion by modifying the exposure parameters: proceeding by trial and error, vary the readout pattern, Ngroups and/or Nints until all pixels at the expected position of the companion are no longer saturated.
JWST ETC Images and Plots
5. Initialize target acquisition (TA) calculations for each of your observations.
All HCI observations will require a science instrument assisted TA procedure—this includes both science and reference PSF observations.
6. Run your TA calculations and examine the output information.
Does any saturation occur?
Saturation should be avoided during target acquisition for optimal performance. If any fully or partially saturated pixels are present in the TA exposure, the ETC will issue a warning. The recommendation is to adjust your exposure parameters (e.g., by decreasing the number of groups) to avoid saturation.
Does your exposure specification allow you to obtain the minimum required SNR for the TA procedure of the instrument mode?
Stage 6 – Select a suitable PSF calibrator
If you have established the need for a PSF reference target according to your PSF calibration strategy designed (see Select a PSF calibration strategy), this section is relevant. Otherwise, you may skip this stage and finalize your observing strategy.
1. Select a PSF reference calibrator with consideration of the following criteria:
Well-known: is the target a known good PSF reference star?
Selecting a reference PSF source that has been previously observed interferometrically/ coronagraphically (or from the ground with adaptive optics) and found to be single, is recommended. "Good references" are usually stars that are not astrophysically contaminated (i.e., without additional astrophysical signal from a debris disk or companion).
- MIRI Coronagraphic Recommended Strategies: choosing a reference PSF target
- NIRCam Coronagraphic Recommended Strategies: Selection of PSF Reference star
- NIRISS AMI Recommended Strategies: Choosing an optimal calibrator for NIRISS AMI
Schedulability: do the visibility windows of the science target and PSF calibrator overlap at the time of the desired observation?
Unless on-orbit experience shows that the need for contemporaneous imaging can be relaxed, the JWST project requires observations of the science target and PSF reference star to be executed together, in a back-to-back sequence of observations.
Users should aim to observe the science and PSF reference observations as close together in time as possible, in order to minimize changes in the PSF and obtain the lowest possible limiting contrast.
Proximity: is the PSF calibrator in relative proximity to the science target?
This is needed in order to limit thermal changes and minimize telescope overheads.
Avoidance of Binary: is the PSF calibrator a single and unresolved source?
This can be addressed by selecting a known good PSF reference star. If the PSF reference star is not well known and/or has not been previously observed with high spatial (e.g., < 0.1") resolution imaging or interferometry, performing further archive checks or seeking another PSF reference star is recommended.
Spectral Type: does the PSF calibrator share the same spectral properties as the science target?
This has a stronger impact at shorter wavelengths and with wider filters. Spectral mismatch may generate extra noise during the process of photometrically rescaling the reference and causing possible under- and over-subtraction of the PSF.
Brightness: is the PSF calibrator similar in magnitude to the science target?
Whenever possible, the use of a reference PSF that is brighter than the science target is recommended because the process of flux rescaling also scales the noise. Selecting a calibrator that is as bright as (or brighter) will help achieve the same signal-to-noise ratio in comparable exposure times.
2. Return to your previous ETC workbook to amend the spectral properties of the reference PSF source and finalize the exposure parameters of your calculations.
Stage 7 – Finalize your observing strategy
In previous stages, you have made a series of choices concerning the content of your observing program—in this stage, you will decide on an observing strategy with which to structure this content. This observing strategy should be designed to mitigate performance degradation and yield the best possible scientific results, with the least possible overheads.
1. Consider the total number of observations you will require for your observing program.
Note that PSF reference observations should be observed using the same telescope optical configuration, so that no wavefront correction should occur between any of the observations.
The inference of the above is that observations in different filters require individual PSF reference observations.
2. At the observation level: consider how you will organize (group) your observations.
Observations that need to be executed together in time should be grouped together in "sequences." Details of these sequences will depend on the science goals of your program.
For all HCI modes, science and PSF reference star observations must be grouped into sequences—the goal is to minimize changes in the optics that might alter the PSF between observations.
Science and reference PSF observations
3. If your sequence of observations involves the use of multiple filters and/or occulters, you should consider following the optimal efficiency scheduling strategy.
With the optimal efficiency strategy, observations for a given target are organized—in each filter and occulter—to minimize the number of rolls and slews. This strategy increases the time between an observation of a target in a given filter and the corresponding reference PSF star observation in the same filter, but it results in more efficient use of the observatory.
With the optimal wavefront stability strategy, standard sequences are executed consecutively in each filter to minimize the chance of any wavefront changes, which comes at the cost of increased number of slews and rolls for the telescope.
Linking too many observations together into a sequence can make the total execution time long, to the point that the observations cannot be scheduled. Therefore, you should seek to strike a balance between efficiency and the pragmatic aspects of scheduling observatory activities.
You may find it instructive to inspect the reports that are generated by running Smart Accounting in APT. These files provide a more detailed breakdown of where various overheads are being charged and will help you understand the tradeoffs in efficiency for the different models. See the articles on the APT Visit Planner and APT Smart Accounting for more information.
4. Do your observations call for a more substantial position angle offset (e.g., 30° offset) on the science target than can be provided instantaneously in a single visibility window?
For instance, this could be needed to recover a part of the scene that would otherwise be blocked by a selected mask. The possibility of such an offset depends strongly on the ecliptic latitude of the target and must be scheduled at a significantly later time. The CVT can be used to help assess this. The special requirements for such an offset should only be used when truly necessary for the science.
5. If your program consists of a set of science targets that are clustered on the sky in close proximity and schedulable at the same time, is it possible to implement the "shared reference survey strategy"?
This is where multiple science targets are paired with an individual PSF observation, in the normal coronagraphic sequence.
Is it possible to incorporate the "self-referenced survey strategy"?
This is under the assumption that some science targets will be for science, but others—those not showing surrounding structure—will be used for PSF reference observations.
6. For all coronagraphic imaging programs: it is highly recommended to perform the standard coronagraphic sequence, or a derivative of it.
7. Do your science goals call for high accuracy astrometry?
If so, perhaps you should obtain NIRCam images for full field astrometry (FFA) in addition to your HCI science data.
Stage 8 – Prepare your proposal in the Astronomers' Proposal Tool
Aside from the steps described in the Getting Started with JWST Proposing roadmap, consider the following particular to HCI:
1. Organize science and PSF calibrator observations into sequences (to be scheduled back-to-back).
You may find it useful to collect all observations that pertain to a particular coronagraphic sequence into a single observation folder, and additional folders for other sequences.
2. Use the PSF reference observations section to indicate which observations produce PSF references and to specify which science observations they should be linked to.
The PSF reference star must be in the same FILTER and SUBARRAY.
You may find it very helpful to use designations in the Name in the APT proposal field (Fixed Targets form) to clearly indicate which targets are intended for science and which are PSF reference stars, as appropriate. These designations will show up in the pull-down menus in other parts of APT, to help you build up your observation sequences. Furthermore, if you have a large number of science targets and PSF stars to keep track of, you may find it useful to do so using the the comment box.
- MIRI Coronagraphic Imaging APT Template: PSF Reference Observations
- NIRCam Coronagraphic Imaging APT Template: PSF Reference Observations
- NIRISS AMI APT Template: PSF reference observations
4. Are NIRCam full frame astrometric (FFA) images needed?
If so, indicate Yes in the Astrometric Confirmation Image parameters template panel and enter the appropriate exposure information for these images.
5. For coronagraphic imaging modes: Do any of your observations require the small grid dithering (SGD) technique?
This selection can be made in the observation template by choosing the appropriate dither type, in the MIRI template, or dither pattern in the NIRCam template.
6. Add any the necessary special requirements:
SEQUENCE OBSERVATIONS... NON-INTERRUPTIBLE to force the Visit Planner to look at the collective schedulability of the entire set.
Note that APT will execute the observations in a Sequence Observations ... Non-interruptible grouping in the order of increasing observation number. If you drag and drop the order of your observations in the APT tree editor, make sure the desired sequence of observations is still in increasing order of observation number. If it is not, edit the observation numbers so that ordering is achieved.
APERTURE PA OFFSET ... for roll-dithered science target observations.
Set the offset angle, or offset angle range between two roll-dithered observations. If a second sequence at a larger PA offset is needed, the Aperture PA Offset ... special requirement must still be set between the two sequences. The cases needing this level of attention to detail should be investigated ahead of planning, with a visibility tool (See Assessing target visibilities and allowed position angles).
APERTURE PA RANGE ... fix the allowed degree range of absolute PA on an observation.
This is only necessary if a known structure around a given target (say a disk or known planet) needs to be positioned to avoid structures in the instrument field of view.
Users should only constrain the requested angles when necessary to support their science goals and that even when an angular constraint is placed, the larger the range that can be allowed the better (from the standpoint of allowing scheduling flexibility). As an exercise, the user can try editing the special requirement that sets the allowed range of angles on the first observation and re-run the Visit Planner to see how the allowed time window changes.
7. Verify your observation set-up.
i. The APT Aladin Viewer can be used to visualize the field of view on the sky for planned JWST observations.
Verify that the position angle(s) and roll dither(s) of an observation/ visit have been specified as intended.
ii. Run the APT visit planner.
Check schedulability of observations, check constraints and see whether guide stars are available to support the observations.
Video tutorials are available for: Reviewing Errors and Warnings
iii. Create Target Confirmation Charts.
This is needed to verify the input target coordinates that will position the telescope in the correct place.
8. Using the Smart Accounting Reports are you able to identify the trade-offs in efficiency (science time/total time) for different observation strategies?
Note: programs that minimize the number of major slews and the number of visits will typically achieve a higher efficiency than programs with large numbers of slews and visits.
Durcan, S., Janson, M., Carson, J. 2016, ApJ, 824, 58
High Contrast Imaging with Spitzer: Constraining the Frequency of Giant Planets out to 1000 AU separations
Konopacky, Q. M., Barman, T. S., Macintosh, B. A., Marois, C., 2013, Science, 339, 1398 (Science link)
Detection of carbon monoxide and water absorption lines in an exoplanet atmosphere
Lafrenière, D. et al. 2007, ApJ, 660, 770L
A New Algorithm for Point-Spread Function Subtraction in High-Contrast Imaging: A Demonstration with Angular Differential Imaging
Perrin, M. D., et al. 2018, Proc. SPIE 10698, 1069809
Updated optical modeling of JWST coronagraph performance contrast, stability, and strategies
Rajan, A., et al. 2015, ApJ 809, L33
Characterizing the Atmospheres of the HR8799 Planets with HST/WFC3
Soummer, R., Pueyo, L., Larkin, J. 2012, ApJ, 755, L28
Detection and Characterization of Exoplanets and Disks Using Projections on Karhunen-Loève Eigenimages