Computational imaging is a fast growing new research field combining new image acquisition technologies with intelligent algorithms. The aim is to extract image information which could not be derived by conventional machine vision. Light field and photometric stereo are two prominent examples for computational imaging.
A light field consists of multiple views of an object obtained from different viewing angles. It can be understood as an extension of conventional stereo systems. This technology uses more than two object views in combination with advanced algorithms and therefore allows for more accurate and robust calculation of depth information.
Instead of using several cameras or a microlens array on the sensor chip (reducing optical resolution), our inline computational imaging approach uses only one multi-linescan camera which simultaneously captures several object views exploiting the relative motion between camera and object. By high-performance computational processing of the light-field data it is possible, e.g. to derive depth information and obtain all-in-focus image with increased signal-to-noise ratio.
This approach provides in addition a photometric stereo setup as not only the viewing angle but also the illumination angle varies due to the relative movement between illumination and object. By analyzing the reflectance properties the slope of the object´s surface as well as information about its material can be obtained.
Both methods complement one another enhancing the overall accuracy and robustness. By choosing the number of viewing angles, the user can decide dynamically on the ratio between speed and accuracy.