The image processing, thus vision, is probably the domain where the parallel processing requirements has occurred very early. In late fifties, first massively parallel processing has been proposed (cf. S. Unger's work in 1958), and effectively realized in beginning of seventies (cf. Illiac IV). Fifty years later a some state-of-art exists. This talk proposes an overview of different processing paradigms inherent to image and vision processing, and of hardware supports for them. It identifies some still open questions in parallel image/vision processing and discuss some future requirements for universal processors (a continuous integration of the structures of today's parallel dedicated computers).