Optical video image compression: a multiplexing method based on the spectral fusion of information

The increasing performances of computer networks and communication systems lead to a huge increase in exchanges of miscellaneous information, including video images with higher and higher resolution. Yet, in many cases, only some parts of the exchanged information are really useful. Thus, in these cases, information can be compressed before being transmitted and/or stored. This compression enables to reduce the processing and transfer time as well as the storage capacity needed for data. Images are of optical nature. Therefore, it is necessary, to be able to use them in these networks and these communication systems, to convert them into digital files before transmitting, computing or storing them. This conversion can require a long computation time or can lead to a bad image quality because of "pixelization" problems. So, it is worth performing some processing on these images (such as compression) directly at the source, before any conversion, by some optical methods. Moreover, there is another advantage of optical solutions compared to digital ones to carry out some image processing operations: their high potential of parallelism (optical processing applies on the whole image in a very short time !). Indeed, with optical solutions it is not necessary to divide the image in several pieces before processing it, in contrast to JPEG digital compression for instance. Thus, for example, it is possible to compute the 2D Fourier transform of a whole image at the speed of light by merely using a convergent lens. In the present work, our aim is to propose and describe a new multiplexing method for video image compression based on an optical image processing: fusion by spectral segmentation, which enables to carry out a fusion that does not destroy the quality of the decompressed output images. The spectral segmentation operation consists in grouping together, in a single spectrum, one part of the spectrum of several images (the spectrum of an image is the result of the 2D Fourier transform of the image). So, because of its selective nature, spectral segmentation allows to perform an image compression in the Fourier domain. With our technique, this segmentation is carried out after a shift (or translation) of the spectrum frequencies of the different images to be compressed. This spectral shift, done before the segmentation, enables to optimize the use of the space-bandwidth product and, thus, to considerably improve the quality of the compressed video images, in comparison to the results got without this shift.