Applications And Realization Of Motion Estimation In Image Stabilization And Matching And Tracking

According to the needs of the practical task, the dissertation deploys the studies on the motion estimation technique and its applications in the electronic image stabilization and image matching and tracking. It also accomplished the hardware design and realization of a high-performance seeker information processor. The main research contents of the dissertation as follows:1. The dissertation introduces the research contents of motion estimation technique and sums up the motion estimation and motion compensation methods in common use, mainly for the electronic image stabilization. It introduces the merits and demerits of mechanical, optical and electronic image stabilization, sums up the overseas and domestic development of electronic image stabilization, introduces the basic theory and the general system framework of electronic image stabilization. Aiming to solving the problem of two dimension image stabilization, it proposes an electronic image stabilization algorithm based on diamond search local-refined gray-coded bit-plane matching motion estimation and kalman filter motion compensation. Based on the study of the one direction and two direction gray projective algorithm, another algorithm based on two directional gray projection algorithm and kalman filter motion compensation is proposed. Furthermore, aiming to solving the problem of the absence of image contents after motion compensation, it proposes a full-frame stabilization algorithm based on image mosaic technique.This algorithm takes the statistical stable center as the estimation of the real stable center and takes it as the benchmark of motion compensation, so as to guarantee the stabilization of the image sequence is carried out according to a more credible stable center.2. In order to solve the problem of image matching and tracking in the condition of translation, rotation and scaling, image registration techniques based on frequency and spatial domain are studied respectively. In the study of frequency domain image registration, it studies the Psuedo-Polar fourier transform and a fast polar fourier transform derived from it, discusses their applications in the image registration, then an improved Fourier-Mellin image registration algorithm is proposed by using the fast polar fourier transform. In the study of spatial domain image registration, it mainly studies the forward additive algorithm, forward compositional algorithm, inverse additive algorithm and inverse compositional algorithm. Then based on the studies of the frequency and spatial domain image registration, aiming to solving the problems that optimization algorithm often suffered, it integrated the improved Fourier-Mellin image registration algorithm based on the fast polar fourier transform with the Levenberg-Marquardt inverse compositional algorithm (LMICA) to gain a frequency and spatial domain combined image registration algorithm, which can exactly estimates the motion parameters between images suffered from translation, rotation and scaling distortion.3. The dissertation sums up several hardware configuration of image processing system and analyses they merits and demerits, then adopts a hybrid hardware configuration based on FPGA and high speed DSP processor as the design scheme and accomplishs a high-performance seeker information processor. FPGA is used to design and realize special image preprocessor, which can accomplish image stabilization and some other global pretreatment during image capturing, which effectively increases the processing speed of the information processor system. It adopts high-performance DSP (TMS320C6203) as the core chip to design the high speed processor and system controller to make them effectively fulfill high data volume and high speed image control and guide information processing.