Mobile robots are frequently engaged in search and rescue tasks. They serve to assist rescue personnel involved in emergency informatics in a variety of ways. One of them involves gathering information about the environment in which they are deployed based on a vision system on-board. The information collected is passed on to remotely located teams. Since the robots often move in an uneven terrain, the camera on-board undergoes vibrations and as a result, the transmitted video tends to be unclear. It is therefore important to stabilize the video and further, it is critical to perform this task in real-time and transmit to remotely located personnel so that appropriate and quick action can be taken. The video stabilization process can be summarized as follows. It consists of identifying static blocks and computing their movements due to a camera shake. In particular, the movement/displacement of static blocks in the current frame with respect to the same static blocks in the previous frame is computed and a displacement vector for each block is obtained. Using this, the global motion is estimated, and the video frames are appropriately compensated.
    In this design, a robust and high-performance approach for video stabilization is developed using Verilog HDL with Matlab. A systolic array for interest point detection and description, key phases in the Speeded-Up Robust Features (SURF) approach for finding point correspondences between two images of a scene and for video stabilization. The overall design is tested using Modelsim with Matlab software.