Adobe’s Content-Aware fill still needs tweaking
Adobe’s AI-powered tool in Photoshop is a clone stamp that replicates or fills in a selected part of a photograph. For example, if a beautifully posed portrait has been photobombed in the background, the Content-Aware Fill can remove the photobomb, replicate the surrounding pixels, then fill in the missing pixels and no one is the wiser. Unfortunately, the results aren’t always accurate if the image is complicated as my results show.
I chose a Kodak photograph from our family photo album to test out the AI. I selected the two figures from the background and used the Content-Aware Fill application to select portions of the image with content sampled from the surrounding parts of the image that would replace the area where the figures once stood.
At first glance, it seemed to be a perfect replica of what was obscured by the foreground figures but on closer inspection, there were odd defects. In place of the baby in the center of the photograph was a leaf and the corner of the small window suspended in mid-air. At the bottom left of the sliding doors were three leaves with jagged L-shaped edges that formed an extension of the step leading up to the sliding doors. This area was replicated and filled the bottom left corner of the photograph.
Despite these defects, the results were impressive, and because I am the baby in the picture who today has no recollection of this garden and house in Los Angeles, I get to fantasize about seeing what my parents saw beyond the edges of a family photograph.
Ginger Liu is the founder of Ginger Media & Entertainment, a Ph.D. practice research student in photography and artificial intelligence, and an author, writer, artist, and filmmaker.
