Research And Implementation Of Digital Video Image Preprocessing Algorithm

With the development of the Internet and the interactive requirement of electronic products, video and image have become an important means of getting information by people. Thus, the image quality requirement on the market is getting higher and high-er. As the intermediate part between the capture system and the post-processing sys-tem, video pre-processing system aims at recovering and improving the quality of the capture image, and provides the post-processing system a high-quality and reliable image. This paper mainly focus on the image denoising, fisheye image correction and image enhancement algorithms, while the first two belong to the field of image reco-vering and the last considered as the image enhancement.Gaussian noise and impulse noise are the two important factors which degrade the image quality. In this paper, denoising algorithms based on image detail protection and multi-decision are introduced to remove the two noise mentioned above separate-ly. As for the detail protection algorithm, standard deviation and sobel operators are used to reduce the Gaussian noise with the consideration of detail protection, while the multi-decision algorithm replace the noise pixel with the most suitable pixel of the filter window based on the multi-decision. Both subjective and objective results of the experiment show the outstanding performance of the impulse denoising algorithm. Moreover, the conception of pipeline structure is used in this paper to the algorithm implementation on FPGA.Besides the influence of the image noise, image distortion is introduced to get more perspective by using fisheye lens, which becomes another factor causing the image quality degradation. Thus, an algorithm based on the projection invariance principle and the geometry properties of radial distortion is proposed. The experimental results show that, compared with straight line fitting method using iteration, the overall qual-ity of the correction image is significantly improved when using the algorithm pro-posed in this paper. Last, an RLE (Run Length Encoding) based LUT (Look up Table) is used to store the mapping relationship between the correction and the distortion image. Then, the RTL design, simulation and hardware synthesis based on FPGA are implemented in this paper based on the mapping interpolation method.After the completion of the image recovering algorithms, an adaptive enhancement algorithm based on dynamic scene estimation is proposed to improve the image qual-ity in various scenes. First, Laplacian operators are used to sharp the image details. Then, four mapping curves are introduced to extend the dynamic range of the image gray-scale values, which improves the contrast of the image. Experimental results show that great effects are got under different scenes by adopting this algorithm, while, at the same time avoiding the distortion caused by the over enhancement and maintaining the original semantic of the image. Last, the conception of pipeline structure is used in this paper to the algorithm implementation on FPGA.The hardware design of all the algorithms proposed in this paper avoids using SDRAM as frame buffer. On the contrary, only SRAM is used as the line buffer to guide the hardware design, which meets the simplification and high efficiency re-quirements of the pre-processing system.