Mean shift procedure can be used in many image/video processing applications for efficiently solving problem of accurate object localization by the computation of center of gravity (CoG) measure. Real-time object localization is a compute-intensive and time-consuming task that is difficult to perform by using only software-based approach. Thus, invariably there is a requirement of accelerating hardware solutions with special emphasis towards efficient and high-performance architectures. This paper presents an embedded framework alongwith required hardware architectures for accurate target localization using CoG computation in the context of object tracking application. Xilinx ML-507 platform has been used to provide a field-programmable gate array (FPGA) implementation of all the required building blocks. The design uses some of the standard intellectual property (IP) elements and an embedded PowerPC processor available in Virtex-5 FX FPGA device of the platform. A high-resolution camera is utilized to capture 640×480 resolution real-time video frames, which are buffered in a DDR2 SDRAM memory of the platform. To compute complex arithmetic operations, the proposed architectures utilize binary logarithm and antilogarithm units with fixed-point datapath. FPGA device utilization shows that the proposed architectures use small number of FPGA slices and a few off-the-shelf FPGA macro elements.