The Core Algorithm Research And System Implementation Of Embedded Image Information Processing System

Based on the research of embedded image information processing system’s core algorithms, this dissertation presents innovational algorithms in three aspects:image amplification, optical character recognition and parallel optimization. The main research content and innovations include:1. Presents a real time image zoom algorithm based on low-cost embedded platforms. Trough divide the original image into sub-regions, then look-up using table mapping method to enlarge each sub-regions, to restore the full image. After building the mapping table data based on bi-cubic interpolation. An improved algorithm is proposed to solve the aliasing problem of connecting parts, by using changeable data in the mapping table.2. In OCR works, the edge of book material would produce surface deformation. These parts would be very hard to recognize by software. In this aspect, an algorithm was demonstrated to solve this kind of problems. The main purpose of the algorithm was to flatten the deformation text information. Through cutting the text into grid, fitting calculation of material deformation function of the curvature, further got the mapping information between raw materials and the flattened materials, then regenerate the flattened information. Through further analysis, a new algorithm was found which can also perform well but didn’t need too much calculation.3. In the implementation of optimization algorithms for embedded platforms with limited computing resources, proposes a hardware-level parallel optimization solutions, describes a search constraints using Lagrange multipliers, multi-process system can improve the performance using high-level synthesis algorithm. The algorithm improves performance by allocate resources according to time slice computing requirement dynamically, thereby reducing the overall computing cycles, improve performance. The algorithm can be integrated with existing high-level algorithm for joint use, and most of the existing tools can be used for further synthesis. From the software level, presents a method of using existing hardware resources, by occupy the free parts to accelerate the software algorithm. Provided a new methodology by showing examples of specific applications for existing hardware.4. According to the algorithms, set up a specific hardware platform and managed to provide a mass production product. Compared with widely used international embedded image processing products, the advantage of algorithms and hardware implementation was shown in performance, effectiveness, power and cost, which also presents very high practical value and business value.