3.4 Image evaluation and pre-selection

In the first QualityFITS step (“QFTIS-in”), all individual CFHTLS input images are used to produce an input .ldac source catalogue (these are FITS binary tables containing an additional table derived from FITS header keywords present in the input images). This catalogue is used first for image quality assessment and subsequently in the astrometric and photometric calibration. QualityFITS also creates for each a image a “confidence map” or weight-map image, which corresponds to the relative inverse variance of each pixel. This weight is derived from the flat-field and mask images produced by the Elixir pipeline (?) at CFHT. This weight map is also used during image combination.

Saturated pixels are also flagged. This saturation cut is much lower than the CCD limit provided in the FITS header (SATURATE FITS keyword) to remove all bright stars and enable star-galaxy separation. During this step we discovered a previously overlooked problem: for some images the highest bit (16th) of the analog-to-digital encoders was stuck to 0, resulting in a 15 bits dynamic in one entire bank of the MegaCam focal place (CCD27 CCD35) versus the normal 16 bits12 . This resulted in different saturation levels across the MegaCam mosaic (32000 instead of 64000). We have modified the .ldac catalogues of the affected exposures outside QFTIS-in with a proper flagging of the two-level saturation without rejecting too many bright sources which are essential for photometric calibration.

After QFTIS-in, images are visually inspected and evaluated through the Youpi image evaluation interface13 . For T0007, grades were preserved from previous survey releases, and only new images were graded. For each image, Youpi provides a link to a QFITS-in web page which summarizes properties of each image. All QFITS-ed images are then graded “A”, “B” or “C”, after a visual inspection of each page (see Fig. 4), paying in particular special attention to the PSF and the seeing over the MegaCam field. Images with grade C are not within the CFHTLS specifications or show serious problems (such as missing CCDs or extensive scattered light over the whole MegaCam field, or obvious telescope guiding problems). “Grade B” represents acceptable images, within the specifications, but for which QFTIS-in revealed minor problems (such as unusual galaxy or stellar counts, or seeing values very close to the upper limit). The grading step is applied to all images which were validated (flag) at CFHT.

Based on the QualityFITS selection a initial selection of images is made by applying the following criteria:

(Note that the number of images with seeing > 1′′ in the CFHTLS survey is actually quite small since the observing requirements set the image quality upper limit at 1.0′′)


Figure 4: The youpi-QualityFITSgrading interface. The top panels show the youpi-QualityFITSmain page which displays the status of QualityFITS evaluations of all images (left), and the QualityFITS evaluation of a user-selected image (right). On the top right panel the grading status for the selected image is displayed. In this case the QualityFITSpage (bottom panel) is loaded by youpiand options to grade a new image or to update the grade of an image are displayed. Users can then examine the data shown on Fig. 48 to 50. The grading interface minimizes the number of actions required to only three clicks per image. The youpidatabase preserves the evaluation history of all images and so it also archives the grades from previous releases. For the T0007 release, only new images have been graded.

Rejected images are removed from the processing pipeline. In subsequent steps, the QFITS-in catalogues of the remaining sample are used to derive the astrometric and photometric calibrations.

The images used in the Wide survey are essentially these images, except that in certain stacks with missing detectors, they are supplemented by supplementary DDT observations mentioned above.

The selection of images used in the T0007 CFHTLS Deep survey starts from a pre-defined set of 85% best-seeing images which has been created in previous releases. It was verified that the seeing on these images (measured using a fit to the PSF with a Moffat function in PSFex) was less than 1.1′′ in u* and less than 1.0′′ in all other bands. In addition, any image which had a field-to-field photometric rescaling (explained in the next section) which was larger than 0.15 mag was also rejected. The resulting list corresponds to our “85%” list. The “25%-best” seeing images were simply constructed from the 25% best seeing images drawn from this list. We verified that the effect of creating such list versus a list from the “best 25%” of all images has minimal effect on the completeness of the final stacks. Despite the name of “85%” and “25%” best seeing images, the lists do not strictly contain 85% and 25% of the total number of exposures as explained in this paragraph.