Editing tools that make images more subjectively real

From New York Times

Mr. Cohen said his tool would produce photos that were closer to the reality that we perceive than a photograph.

“We’re assembling what’s really there — just not from one-hundredth of a second, but longer,” he said.

“Think of an axis from the purely objective to the purely subjective,” he said. “At one end is a photograph, a recording of what really took place. At the other end is our internal experience of an external event. There’s some place that is a little bit subjective. It’s not quite real. But if you and I looked at it, we would agree on it.”

This story is about tools that let users edit images to make them look more like we remember them. For example, one tool will combine multiple pictures of a group of people so that no one is blinking—even though in each individual picture at least one person is blinking. An effect I've noticed is that telephone and electric wires seem much more prominent in pictures than they do in reality. When we look at a scene, we tend to ignore foreground telephone wires. But when those wires appear on a photograph, they're a lot harder to ignore. these tools attempt to make pictures more like we see them subjectively.

The top picture is the original. The second one stretches the scene to fill a larger area, e.g., for use as a screen saver, but it doesn't stretch the dog's face. You'll notice that the trees behind the dog look the same in the two pictures, but the trees to the left and the right are wider in the second picture. The point is that the subject of interest, the dog's face, is retained as we remember it. The other parts of the image are stretched but in ways we don't notice or don't care about.

The way the stretching works is by finding a path of pixels from the top to the bottom where adjacent pixels are the same, or as closest to the same as any other path. Then that line is replicated, adding a pixel width to the picture. This is repeated until the picture is stretched as much as one likes. Presumably the software checks to ensure that it doesn't repeatedly insert pixels in the same places over and over.