In this paper we introduce a new object recognition method for detecting objects of arbitrary shapes based on a combination of the Fast Hough Transform (FHT) and the Dual-Point Generalized Hough Transform (DPGHT). We validate the proposed approach on medical images (cryosection, US, MRI) and describe its basic features. The proposed algorithm is robust, low memory consumptive and utilizes the advantages of the both FHT and DPGHT that makes it a good solution for medical object recognition applications.