Research And Implementation Of Fisheye Camera Image Processing

Fisheye camera has the characteristics of wide Angle, large depth of field, and it has been more widely used in the field of photogrammetry and omni-directional vision. Because of its particular construction,the application of fisheye camera has been gradually extended to public security, national defense and aerospace monitoring and other special areas. In these special applications, with the spatial resolution of the camera from the standard definition gradually ascending to 1080 p and 4k, and the frame rate by 25 frames per second gradually rising to 60 frames per second, distortion image preprocessing and real-time correction has become the choke point of its application, and the related algorithm and hardware implementation architecture has become the key problem of high-speed implementation.With the support of cooperative 863 unit, this paper has carried the thorough research on the fisheye camera image processing flow, bad pixel correction and fisheye distortion correction. The main content of the thesis has three aspects. Firstly, in view of the fisheye images which are the output of the cooperation unit ?s optical system, this paper designs the efficient bad point correction algorithm. the camera output image containing a large number of texture information, The existing bad pixel correction algorithm is effective in flat areas of the image.,but it is not very effective on the edge of the image,especially complex edge,and it is unable to filter out the bad pixel on the edge of the image, or it leads to fuzzy edges after processing. This paper puts forward a kind of bad pixel detection and correction algorithms which is based on the minimum value filter in the gradient.During the bad pixel detection, The minimum value in the gradient is used to indicate the image edge direction, and detect the image along the edge.During the bad pixel correction, the adaptive correction method is based on the rule of the edge gradient,so that the output image of this processing can restore the texture of the target image.Secondly,based on the the bad pixel correction algorithm in this paper, this paper designs the efficient processing hardware structure, and through a simple add and subtract and shift processing, this algorithm obtains high performance correction effect.On the other hand, based on the spherical perspective projection method, this paper designs the efficient hardware processing architecture. In the process of fisheye image correction, the divider is too complex,and according to the laws of the fisheye image output coordinate alternation,this paper converts division into multiplication and reduces the complexity of the algorithm, and improves its real-time performance. In the process of image buffer processing,it is a problem to read the rectified the fisheye image. Based on the characteristics of the ranks of the fisheye image distortion, this paper proposes the methods of how to read the address based on the row pointer and column pointer. Each reading address after correction are made by the previous address with the corresponding offset,so that it is simple and direct to read the SRAM and improve the overall processing speed, save the hardware resources.This paper completes the complete process of the the cooperation unit?s camera, designs and implements the corresponding hardware structure of the bad pixel correction algorithm, furtherly this paper completes the fisheye correction algorithm processing architecture,so as to make it become a part of the camera image processing. The hardware implementation results show that the proposed algorithm can filter the bad pixels of the cooperation unit camera?simage effectively. This architecture can correct the image output of fisheye camera in real time. It can meet the requirements of the image real-time processing of the fisheye camera, and provide a high reference value and application value.