Color Image Denoising And Video Capture Software Design

In the video surveillance system, video capture card is the core of the system. The images will be affected by external factors during their collecting, so that image quality was descreased, and important image detail was lost, and then it will be more difficult for the subsequent image processing, such as image segmentation, integration and recognition, etc. Therefore, the foremost task was to isolate and remove noise before the further image processing. The wavelet is a good tool for signal analysis, which can effectively isolate and remove noise, and the research method has been mature. However, wavelet analysis has a point of singularity, so it could not use geometrical features of data to get sparse representation of the signal which is two-dimensional or higher dimension. Curvelet multi-scale analysis method has strong directional for the signal with curve singularity, and is favorable to portray image edge, so it has made up deficiencies in the wavelet analysis method. Therefore, the research of curvelet transform is necessary.This paper deeply studied color image denoising method based on wavelet and Curvelet analysis, and designed software development kit of video compression card to implement video surveillance, then denoised the captured images. The main contents are as follow:(1) The color space characteristics of image and their influence for color image denosing were analyzed, and the paper selected YCbCr color space to denoising, which can more effectively remove noise without sacrificing image quality.(2) Wavelet threshold denoising algorithm was studied. On this basis, the method of YCbCr color space BayesShrink and MapShrink threshold denoising were adoped. This method can not only better maintain the color information, but also be able to obtain a higher PSNR (Peak Signal to Noise Ratio).(3) Curvelet multi-scale analysis method was researched, and YCbCr color space threshold denoising based on Curvelet transform was proposed. This method constructed an appropriate threshold and improved threshold function. In addition to better maintain the image color information, the method can retain more image details, which can greatly reduce fuzzy of the image edge generated by the wavelet approach and increase the PSNR.(4) The video capture scheme based on SDK was designed. C++ programs of API function libraries and applications of MPEG-4 video compression card based on PC/104plus architecture were designed by using VC++6.0. The real-time video preview and the capture of surveillance images were achieved by the technology of multimedia processing based on DirectShow. The proposed method was adoped to denoise captured image, which can effectively remove the noise and obtain a higher PSNR.(5) The video capture scheme based on DirectShow was designed. Under the DirectShow architecture, the application software of MPEG-4 video compression card was achieved by programming.The experimental results in this paper prove that the denoising method is effective, and the the software development kits designed for video capture implement the functions of MPEG-4 video compression card and are stable.