Some UVIS images may contain features that are not direct images of astronomical sources. The causes of these features include multiple reflections between optical surfaces (detector, filters, and windows) of light from the astronomical scene, scattered light from bright sources outside the detector FOV, light from the bright Earth that is scattered in the OTA, and electronic cross talk between readout amplifiers. In general, these artifacts are not calibrated and cannot be removed by the WFC3 pipeline.
Examples of the different kinds of image anomalies can be found on the WFC3 website as well as in WFC3 ISR 2017-22. 

5.5.1 Ghosts

Ghosts appear as images of the pupil formed from the light of a bright target in, or near, the UVIS detector FOV. The target light is scattered twice (or more) by optical surfaces forming one (or more) out of focus images. The separation of the ghost from the source depends upon the separation of the scattering surfaces and the angle of scattering. Filter ghosts are formed by scattering of the near-normal source light at the surfaces of a filter and, as such, are found close to, or overlapping, the source image. Further details concerning these ghosts may be found in WFC3 ISR 2007-09. Some examples of the ghosts can be seen in Figure 5.10.

Scattering between the UVIS CCD and either the detector, or dewar window forms so-called ‘optical’ ghosts. These ghosts are separated by ~80" from their source generally in pairs (a "figure eight"; see the right panel of Figure 5.10). Further details may be found in the following WFC3 ISRs; note that documents released before 2009 were based on ground test data. The reports include ISR 2001-17 (ghost modelling), ISR 2004-04 (initial filter ghosts, most fixed before flight), ISR 2007-09 (filter ghosts after filter replacements with overall performance of those replacement filters in ISR 2007-01), ISR 2007-21 (channel glint in spare UVIS detector ground-testing images, not present in on-orbit images) and ISR 2011-16 ("figure eight" ghosts). 

5.5.2 Dragon’s Breath and Scattered Light

Bright sources immediately outside the image frame on WFC3/UVIS can create (mostly minor) light anomalies in the form of spurious signal redirected into the field of view and affecting the image (see Figure 5.11). Dragon’s Breath (when the light originates from the edge) or Scattered Light (when the light is somewhat displaced from the edge) affects about 12% of a subset of images (WFC3 ISR 2017-02). It occurs when a bright source lies closer than approximately 30’’ outside the field of view, and peaks when the neighboring source has a V band magnitude around 11.9 or 14.6 (interactive visualization of known dragon's breath sources). The cause for the scattering along the optical path is not completely understood. Most cases of Dragon’s Breath and Scattered Light are low-level features and will not adversely affect a science image. Note that dithering, done by the vast majority of observers, results in either a shift or sometimes complete disappearance of the feature.

5.5.3 Crosstalk

Generally whenever two or more quadrants of an electronic device are read out simultaneously, there is a chance of crosstalk (Janesick 2001) and in fact, both the UVIS and IR channels in WFC3 exhibit some very low-level crosstalk (CT). In the UVIS, amps A+B and amps C+D are coupled, thus bright sources in A (or B, C, D) will give rise to CT in B (or A, D, C) respectively. In the IR channel, the CT is also a low-level mirror image although in this case, the coupled amps are 1+2 (upper left and lower left, when image is displayed with x=1,y=1 at lower left) and 3+4 (upper right and lower right). In both channels, the CT appears as a negative image; thus, these electronically induced features are unlikely to be confused with e.g. optical ghosts.

Figure 5.12 shows a sample UVIS image with CT in both amplifiers due to multiple bright sources. The UVIS CT is linear, negative, and appears at the level of ~10^-4 to 10^-5 of the source. Specifically, in full-frame, unbinned UVIS readouts, the CT level is ~2 × 10^-4 of the source when the target is in quadrants A or C and about 8 × 10^-5 when the target is in quadrants B or D (WFC3 ISR 2009-03).

Crosstalk in the UVIS channel only occurs in the chip containing the target, it does not cross between chips. To within the errors, the CT due to hot pixels and cosmic rays is the same as for point or extended sources. Dithering of observations can help mitigate the low-level effects of CT: the mirror image nature of the CT moves the features in a direction opposite to the target motion, i.e., they will appear to be transients and thus be removed during the drizzling procedure.

A standalone IDL procedure is available for correcting UVIS data for crosstalk (the automated pipeline does not remove the effect). Pixels are effectively corrected to a mean which is well within 1-sigma of the mean of surrounding pixels. The code is described in WFC3 ISR 2012-02 and the software (IDL and new python version) are available for use. Future versions of the python code will be distributed via the WFC3 software Library.

Figure 5.13 shows a subsection of amplifier B from the image in Figure 5.12 and the corresponding crosstalk corrected image.