First Focus Invariant Images

First Focus Invariant Images

In early 1995, Dr. Joseph van der Gracht of the Army Research Laboratory (ARL), performed the first laboratory proof-of-concept experiment of focus-invariant imaging. He demonstrated that by using a special-purpose optical mask an incoherent optical imaging system can be insensitve to focus. After linear digital filtering of the recorded imagery, the optical/digital system produces images with a very large depth of field, relative to a traditional imaging system, while preserving the spatial resolution and light gathering power of the traditional system.

Use the buttons to view the traditonal, raw focus invariant, and filtered focus invariant images

The images from the traditional system are of a planar square and circle black and white target. The target location was moved in order to produce images that are infocus, have mild misfocus, and have extreme misfocus. The mild misfocus of the center image is roughly ten times that of the Hopkins criteria for misfocus. The extreme misfocus is roughly thirty times that of the Hopkins criteria for misfocus. At extreme misfocus all targets are unrecognizable.

Intermediate images are raw sampled images from focus-invariant systems. These system are physically a traditional optical imaging system modified by a special-purpose optical element or mask. This mask theoretically absorbs no optical power. An image of the optical mask is found here . The intermediate images are not as sharp and clear as the infocus image from the traditional system. But, the intermediate images are nearly the same for each value of misfocus, especially when compared to the changes in the images from the traditional system with misfocus. Digital filtering of these intermediate images restores the spatial resolution to that of the traditional system.

Linear digital filtering of the intermediate images produces the final focus-invariant imagery. Only a single linear digital filter was used to restore all parts of each of the images. The spatial resolution of the geometrically extreme misfocus image is nearly the same as that of the geometrically infocus image and all of the focus-invariant images are comparable to the infocus image produced from the traditional system. For this experiment an off-the-shelf analogue video camera with roughly six bits of dynamic range was used.