Barley, a dog, leaps after a yellow seahorse toy that has been thrown. Both Barley and the toy show signs of motion blur, as well as a more subtle halo of “unenhancement blur.” Picking up on last week’s theme of “photos of my dog that are blurry,” I revisited this older photo recently and noticed something interesting. Barley and her seahorse Stella are blurry - this is unsurprising, since both are in motion. However, look carefully at the areas around them both, in particular the carpet. The ‘radius’ of blurring caused directly by Barley’s motion looks to be within 16 pixels or so, as judged from tracking the blur on her collar. The carpet in the background around Barley, however, is blurry to out to four or five times that distance. Go ahead and zoom in, see if you can spot the weird halo effect in the carpet around Barley. It’s at though parts of the camera had more time to get its act together, but couldn’t quite figure out what it was looking as it got closer to the action. I suspect this is a tell regarding just how much post-processing a camera phone does before presenting you with an image. I bet this camera took a lengthy burst of captures from its CMOS sensor, did an analysis of which regions were sufficiently stable to keep, and stitched this photo together from a mix of high-data/low-speed sampling from the periphery and low-data/high-speed sampling from the objects in motion. That wider “blur” is only visible because the texture of the carpet is about the same resolution as the background chromatic error, so even with a perfectly still camera, resolving that texture requires a longer sample time. That’s just speculation on my part, of course, but it goes to show just how much invisible artifice already goes into digital photography, even before we consider our looming nightmare of ubiquitous transformer-architecture cameraphone editing.
Read more →