Optical Computing Techniques for Image/Video Compression

The advantage of optics is its capability of providing highly parallel operations in a three-dimensional space. In this paper, we propose optical architectures to execute various image compression techniques. We optically implement the following compression techniques: Transform coding • vector quantization • video coding We show many generally used transform coding methods, for example, the cosine transform, can be implemented by a simple optical system. The transform coding can be carried out in constant time. Most of this paper is concerned with an innovative optical system for vector quantization using holographic associative matching. Limitations of conventional vector quantization schemes are caused by a large number of sequential searches through a large vector space. Holographic associative matching provided by multiple exposure holograms can offer advantageous techniques for vector-quantization-based compression schemes. Photorefractive crystals, which provide high-density recording in real time, are used as our holographic media. The reconstruction alphabet can be dynamically constructed through training or stored in the photorefractive crystal in advance. Encoding a new vector can be carried out by holographic associative matching in constant time. An extension to interframe coding using optical block matching methods is also discussed. Most of the results in this paper were previously presented in our earlier paper [1]. © 1994 IEEE

Duke Scholars

Author John H. Reif Computer Science