VIP v1.1.0

• Major revamp of the package:

1. creation of a 'fm' module for all forward-modeling related routines (disk model creation, negfc, fake companion and fake disk injections);
2. creation of a 'psfsub' module including all PSF modeling and subtraction post-processing algorithms;
3. creation of a 'invprob' module including andromeda, and to host future inverse-problem approach based algorithms;
4. fixed relative imports in package to match the new structure.

• Additional automatic tests for stim map, fake disk injection, (de)rotation, image scaling
• Improvement and bug correction of the phase function for the disk scattered light forward model tool
• New bad pixel correction algorithm based on kernel-interpolation;
• Updated write_fits to also allow for multiple-extension fits files to be written;
• Added border_mode option to all recentering functions;
• half_res_y option added to sigma clipping algorithms (i.e. made it compatible with image-slicer IFUs);
• Photutils and numpy deprecation warning fixes;
• Routine descriptions adapted for compatibility with sphinx documentation;
• Updated matplotlib version requirement to avoid bug with the latest mpl version (as of 02/2022);
• Default start method for multiprocessing set manually to 'fork', since it was changed to 'spawn' as default starting from python 3.8 (leading to bugs in all routines using multiprocessing);
• Updated default flux range explored by grid search negfc.

VIP v1.0.3

• Addition of new image (de)rotation options to mitigate Gibbs artefacts when rotating an image with sharp signal transitions
• Fix for injection of shifted psf images with FFT-based method
• Fix for multiprocessing in frame differencing function
• Fix for multiprocessing (de)rotation of images
• Addition of some automatic tests to the pytest suite
• Fix for automatic SDI test
• Modification of the automatic RDI test
• Modification of the default waiting time to open ds9 window, to avoid a non-response bug.

VIP v1.0.2

Two major changes in default VIP behaviour:

1. Change of default method for image shift, scaling and rotation throughout VIP: switched from opencv interpolation-based methods to FFT-based methods (imlib='vip-fft').
2. Change of VIP convention regarding center of even-sized images: dim/2 (zero-based indexing) instead of dim/2-0.5. For odd-size images, the convention is identical. Now, in either cases the centroid is assumed to fall exactly on a pixel.

Readme and andromeda routines were adapted to reflect the above change in convention.

VIP v1.0.1

• [preproc] Addition of scale_fft, an FFT-based image scaling routine, which can be used in frame_px_resampling, cube_px_resampling, cube_rescaling_wavelengths and frame_rescaling by setting imlib='vip-fft'.
• [preproc] _frame_rescaling renamed into frame_rescaling (i.e. made callable from outside rescaling.py).
• Fix for opencv implicit dependencies in throughput, contrast curve, speckle noise estimation, cube recentering via speckles, STIM and inverse STIM map functions.
• Added dtype=object for ragged arrays to avoid deprecation warnings (in snr and utils_pca functions).
• Fix for 3D numpy arrays opened through vip_ds9.

VIP v1.0.0

Major update:

• requirements built with pip-compile (pinned versions of packages)
• New module: specfit
• imlib and interpolation now consistently provided to all ADI-related algorithms (used for rotation)
• [negfc] Added routine to estimate residual speckle noise uncertainty by injection of fake companions
• [negfc] New log-likelihood expression for MCMC-NEGFC
• [nmf] NMF full-frame compatible with masking of negative values
• [nmf] NMF annular
• [pca] New options for PCA routines
• [pca] Updated default behaviour for pca_annular with reference cube such that R+ADI can be performed depending on delta_rot value
• [preproc/derotation] added FFT-based rotation algorithm (Larkin et al. 1997)
• [preproc/rescaling] New automatic algorithm to determine optimal scaling factor between images at different wavelengths.
• [preproc/subsampling] NaN-compatible collapse functions (e.g. np.nanmedian instead of np.median)
• [var/filters] filters can now interpolate NaN values (a la astropy)
• minor bug fixes
• deprecation fixes

VIP v0.9.11

Changes since VIP v0.9.10:

• Improved compute_binary_map function
• Updated docstrings in several functions
• Fixed call to deprecated normpdf (matplotlib older version)
• Fixed bug in cube_detect_badfr_ellipticity (to match new photutils API)
• Fixed how delta_rot parameter was handled in xloci and pca_annular functions
• Fixed negfc.firstguess function

VIP v0.9.10

Changes since VIP v0.9.9 :

• New SVDecomposer class, svd module was cleaned up.
• Lots of improvements to vip.pca.pca function for ADI, ADI+RDI or AmSDI datasets. It allows the computation of a grid of results when ncomp is a tuple, the possibility to choose a desired CEVR, and incremental PCA for big arrays. Updated docstrings.
• Improvements to vip.pca.pca_annular function for ADI, ADI+RDI or AmSDI datasets. Updated docstrings.
• Improvements to vip.preproc.cube_recenter_dft_upsampling function.
• Defaults changed in vip.var.fit_2d module.
• Added test modules: test_pipeline_adi, test_pipeline_sdi, test_pipeline_rdi, test_metrics_detection and test_metrics_snr.
• Module vip.metrics.snr_source has been cleaned up. Function snr computed the STD of background apertures with the ddof=1 option.
• Other minor fixes and improvements.

VIP v0.9.9

Thanks to @r4lv, @VChristiaens, @jmilou, @ChrisDelaX who made some of the contributions below. Changes since VIP v0.9.8 :

• Dropped Python 2.7 compatibility.
• Removed plotting module. VIP now imports the package HCIplot.
• Improved test suite (pipeline for both ADI and ADI+mSDI).
• Improved detection function.
• New options for the S/N computation.
• Better handling of versions in requirements file.
• Bug-fix in PCA for 4D arrays.
• Bug-fix phase_function module.
• Bug-fix in get_eigenvectors when ncomp = None.
• Improved OOP interface (HCIDataset, HCIFrame and HCI).
• Improvements to the vip_hci.metrics.roc module.
• Coverage reports with codecov.io.
• Updated ANDROMEDA algorithm.
• Added the stim module (detection map) in metrics.
• Improved documentation in several functions/classes.

VIP v0.9.8

Changes since VIP v0.9.7:

• New class ScatteredLightDisk by @jmilou for generating synthetic discs. This code is based on a light version of the GRATER tool (GRenoble RAdiative TransfER, Augereau et al. 1999)
• New functions cube_inject_fakedisk and cube_inject_trace
• New function cube_shift
• Bug fixed in snrmap function
• Bug fixed in cube_detect_badfr_pxstats (due to Pandas API change)
• Improvements in HCIPostProcAlgo and HCIDataset classes (for new OOP API)
• Improvements in normalize_psf function
• Improvements in cube_rescaling_wavelengths and _find_indices_sdi functions
• Improvements in median_sub function
• Improvements in compute_binary_map function
• Redundant function in shapes module were removed. The signatures of several functions were unified
• Changed defaults for xloci least-square model PSF subtraction
• Updated ANDROMEDA algorithm
• Support for Python 3.7
• Improved pytest coverage
• Improved documentation

VIP v0.9.7

Changes since VIP v0.9.6:

• Several model PSF subtraction algorithms are now compatible with ASDI datasets.
• New metrics subpackage with a ROC class.
• Refactoring of multi-processing logic.
• Improvements on the ANDROMEDA implementation.
• Various bugs were fixed.
• Code cleanup.