IPP Overview

The Pan-STARRS Image Processing Pipeline (the IPP) is responsible for performing the primary image analysis for the Pan-STARRS project. This includes: correction of the instrumental signature for individual images; detection and basic classification of astronomical objects in images; astrometric and photometric calibration of the images; geometric tranformation and combination of sets of images into cleaned stacks; PSF matching with reference images and image differencing; and construction of a high-quality reference image for all regions observed.

The IPP consists of a collection of software tools for performing these analyses, the data management infrastructures, and the necessary tools for automating the processing and monitoring the results. The links at left discuss in some detail these different portions of the IPP. Note that the IPP is designed with a very flexible configuration system which can allow for its use with mosaic images from a range of astronomical telescopes. To date, most of the testing has been performed with images from the CFH12K, MegaPrime, and Suprime mosaics.

The IPP is designed to support both the Pan-STARRS single prototype telescope, PS-1, or the full Pan-STARRS telescope array, PS-4. With PS-4, every observation will consist of four overlapping images of the same portion of the sky. For PS-1 some of the observing will be designed to mimic this data product with sequential observations of the same patch of the sky. Other observing under PS-1 (eg, the AP Survey) will have a difference cadence, and may not consist of more than a single image for a patch of the sky. The IPP analysis strategy is designed to support either of these modes.

The IPP analysis is divided into the follow 6 major stages:
  • Phase 0 : Image Registration In this stage, the image is added to the IPP database table of raw images; the image type (calibration, science, etc) is determined; basic statistics are extracted from the headers and made available from the database.
  • Phase 1 : Mosaic Pre-Calibration This analysis step is only applied to mosaic images. A sample of bright stars in the image are examined an initial astrometric calibration for the full field is determined with modest accuracy. If OTA guide stars are available, these are used, otherwise bright stars are identified from the astrometric reference catalogs.
  • Phase 2 : Image Analysis This step consists of the primary analysis which is performed on each individual exposure: detrend correction, object detection, astrometric and photometric calibration. The depth and detail of the object analysis depends on the observating strategy, but is normally fairly shallow for images which are obtained with a stacking strategy. This analysis is performed independently on individual chips of a mosaic camera exposure.
  • Phase 3 : Mosaic Calibration This analysis is only performed on mosaic camera data. The astrometric and photometric results from Phase 2 for the different chips of a mosaic camera are assembled and used to construct improved astrometric and photometric calibrations which account for variations of the transparency as a function of position, and using a single astrometric model for the optical distortion.
  • Phase 4 Sum : Image Stacking This analysis is performed on a group of images which observe approximately the same patch of the sky. The images are all geometrically warped to a common pixel grid, tied to a celestial coordinate system. The images are combined in a robust way to reject bad input data: dead pixels, hot pixel, cosmic rays, etc. The output pixel grid is chosen to match one of the images in the reference image collection, which we call the Static Sky. This analysis is performed independent for each static sky image (Static Sky Cell).
  • Phase 4 Delta : Image Difference The stacked image and the Static Sky reference image are PSF-matched by determining an appropriate transformation kernel. The reference image is subtracted from the stacked image and transient objects (moving or varying) are identified in the resulting difference image.
  • Phase 5 : Static Sky Update The transient objects detected in Phase 4 Delta are excised from the input stack images and a new stack is constructed. This stack is then used to improve the signal-to-noise in the Static Sky reference image.
  • Static Sky Analysis In addition to the analysis performed on each image or group of images, the Static Sky images are regularly analysed with detailed object analysis.
  • Detrend Image Analysis note: this probably needs its own section.
discuss the parallel processing strategy