Few papers discuss the properties and applications of the powerful Zoran Vector Signal Processor (VSP). In addition to giving an FFT-based autocorrelation subroutine, the authors provide some specific details of the use of the vector instructions and outline some limitations of the fixed-point Zoran processor. The vector DSP is only a means to achieve the main goal of character pattern matching for facsimile machines and photocopiers. The authors briefly but clearly review the algorithms used, which include matched filtering, statistical matching, and cosine transform. They emphasize peculiarities of the cosine transform in the light of a VSP implementation. The quantization of cosine transform coefficients for image compression purposes is briefly addressed. Some minor ambiguities exist, such as the statement that fast convolution can be used for fast correlation if a bit reversal is applied instead of time reversal. This work is a good practical engineering paper.