Some new results in hybrid acousto-optic processing

A fully programmable hybrid acousto-optic convolving processor has been designed for transversal and recursive filtering applications. This system is used for specific signal processing tasks. Its performance (computing time, available precision, noise limitations) is characterized for each task. The main results are reported, with emphasis on transformations pertaining to the image processing field (bidimensional convolution, gradient and Laplacian operators, discrete Fourier transform, discrete cosine transform). Future applications are discussed, including the implementation of algorithms of the JPEG and MPEG classes and work toward the design of hybrid acousto-optic neural networks.<>