The use of image processing in tracking

Digital image processing is a very widely applicable technique in many different areas. Images are not restricted to photographs, drawings and TV-pictures but include any sets of digitally encoded figure points. An important step in an image processing sequence is extraction of features. This involves recognition of patterns of all sorts, e.g. blobs, clusters, line segments and curves. In future high-luminosity high-energy physics experiments, information from difficult kinds of detectors, due to the high resolution and fine granularity, can be considered like images. Because of the large amounts of data involved, image processing techniques can be employed to great profit. A common task in particle physics experiments is that of recognizing tracks. It can be done to different degrees of refinements - presence or absence of tracks for fast triggering purpose, track parameter calculation for higher level triggering and final track reconstruction for data symmetry tapes. In the following, one particular method will be discussed namely the Hough transform. We have implemented, as part of the LAA project, this method in a special-purpose processor. We discuss implementation, results and extensions towards higher degrees of parallelism. (orig.)