Problems related to reflection of some algorithms of image processing onto multiprocessor systems are investigated. In particular, algorithms of allocation of contours, such as the gradient method and the combinatorial method (the method of the threshold gradient), are considered. In view of complexity of the algorithm of searching objects on an image, we focus our attention on the parallel variant of only one stage of this algorithm. Namely, we investigate the stage of working with a search tree. Thereby we suppose that architecture containing a switchboard or the multiprocessor Cell developed by IBM Corporation is used in all cases. Data allocation, all data transfers, etc. are described. The formulas for acceleration coefficients are obtained.