基于GPU并行计算机的图像增强技术研究【毕业论文】.doc

本科毕业论文 （20 届） 基于GPU并行计算机的图像增强技术研究 专业：计算机科学与技术 摘要 随着电子计算机技术的进步，计算机图像增强近几年得到飞跃发展，已经成功的应用于几乎与所有与成像有关的领域，并正发挥着其重要作用。图像增强是指按特定的需要采用特定方法突出图像中的某些信息，同时削弱或去除无关信息，以改善图像的视觉效果，将原图转换成一种更适合人眼观察和机器进行分析处理的形式的图像处理方法。在图像处理中，图像增强技术对于提高图像的质量起着其重要作用。传统的图像增强方法有空间域和频率域。空域法主要是对图像中的各个像素点进行操作，如直方图均衡化和锐化处理、空域滤波等；而频域法是在图像的某个变换域内对图像进行操作，修改变换后的系数，例如傅立叶变换等，然后再进行反变换得到处理后的图像,其算法主要包括多种滤波算法和基于小波、Curvelet、Contourlet等变换算法。 本文还介绍了几种图像增强方法；基于直方图均衡化的图像增强算法，基于空间域的图像增强算法，基于频率域的图像增强算法，基于小波变换的图像增强算法。针对一些图像的分辨率高，增强算法的运行时间较长，分析并利用CUDA GPU 的并行处理特性，提出了一种基于CUDA平台的图像增强并行加速算法，它利用GPU的并行运算功能来完成塔型算法的分解和重建环节中的大量卷积运算。实验表明，该方法能显著提高计算速度，随着图像分辨率的增加，加速比也增加，当分辨率达到2000×2000以上时，加速比可达到百倍之多。 【关键词】 图像增强, GPU, 直方图均衡化,CUDA GPU-based Parallel Computer for Image Enhancement Technology Abstract With the development of the computer technology, the computer-image enhancement has been developed rapidly in recent years. And it has been successfully applied to almost all areas of imaging-related, and is playing an important role. Image enhancement is defined as the need to adopt a particular method of striking images in some of the particular information. In order to improve the image of the visual effects, the image enhancement will at the same time weakens or removes the irrelevant information. It will change the original image into a more suitable one which is a new way to analysis and process the image. In the ways of image processing methods, image enhancement technology plays an important role in improving the quality of the image. The traditional image enhancement methods are spatial and frequency in domain. Airspace is mainly in an image point of each pixel operations such as histogram equalization and sharpening and partial filtering; while the frequency domain method is a transformation which is to operate within the image, and then modify the transformed coefficients. Such as the Fourier transform, which is an anti-transform of the processed image. It contains a variety of filtering algorithms and wavelet, Cu