Computational imaging and machine vision improve image processing

Computational imaging (CI) redefines the design of traditional image capture for machine vision applications and offers cost savings and flexibility for manufacturers.

By AIA March 12, 2020

Most modern digital cameras have a lot in common with the film camera you took on vacation in the 90s. But a new technology called “computational imaging” (CI) redefines the design of traditional image capture. The complex technology of computational imaging solves many of the issues that plague machine vision.

CI processing software provides ultra-resolution color, enhanced contrast, and extended depth of field. It can also generate 3D surface data, remove glare, and embed multispectral data into a single image. CI offers significant cost savings and flexibility over capturing similar quality images from other high-quality cameras.

What is computational imaging?

Computational imaging takes a new approach to image capture. With a traditional camera, you have a single exposure that’s taken with fixed optics and lighting. If the image produced for a machine vision application with a traditional camera isn’t up to par, pre-processing software is the only way to enhance the images before processing.

Lens control systems paired with programmable lighting allow for high-quality image capture thanks to the ability to set application-specific parameters. The light source’s direction, intensity, wavelength, polarization, and focus can be changed on the fly.

One particularly helpful advantage of CI is that it starts with a sequence of images that are combined into one. The composite image is much sharper. As well, complete color information can be captured at the full pixel resolution of the imager. With CI, a single camera can be used to capture both monochrome and color images.

Common computational imaging methods for machine vision

Many CI techniques are used for machine vision. Photometric stereo CI separates an object’s shape from its 2D texture or surface coloring. This technique allows a machine vision system to detect 3D surface features and imperfections quickly. Another advantage of photometric stereo CI is that it can remove glare from highly-reflective parts.

Multispectral imaging combines the ultraviolet, visible, and infrared spectra to enhance images with maximum contrast. It also works well with highly-reflective materials. The various light wavelengths help machine vision in different ways. To clarify, visible light offers a typical visual profile, near-infrared light helps find defects, and UV light lets machine vision see dye markings.