【作者】 潘桂芳；

【导师】 胡青泥；

【作者基本信息】 大连理工大学 ， 机械设计及理论， 2005， 硕士

【摘要】 图像的感兴趣区域ROI(Region of Interest)压缩编码技术,即在图像中的感兴趣区域(ROI)进行无损或近无损压缩,而在其它区域采用有损压缩,从而既可以获得所期望的高质量的图像信息,又保持了较高的压缩比,很好地解决了压缩比和图像质量之间的矛盾。 感兴趣区域编码是JPEG2000的重要功能之一。ROI的提取是实现该功能的首要问题。图像分割的技术可以用来作为提取ROI的基本手段。本文针对多幅具有相同目标的彩色图像,实现了ROI自动提取。提取的方法为:首先从一幅图像中指定包含目标的矩形区域作为ROI,提取和描述ROI特征。对其它图像进行边缘检测,生成包含边缘的矩形区域作为ROI检索区域。运用已记录的ROI特征与其它图像中的ROI检索区域进行匹配,得到相似度较大的区域为ROI,完成ROI的自动提取。针对本文的研究对象,本文做了以下工作: 1)彩色图像的ROI自动提取。现有的ROI提取算法大部分针对灰度图像,这些算法对彩色图像的ROI提取并不能直接适用,本文在研究现有算法的基础上制定了对包含相同目标的多幅彩色图像的ROI提取方法,实现了对这类彩色图像的ROI自动提取。 2)彩色图像包含的颜色信息丰富,单独采用一种图像分割方法,很难完整的提取目标。本文在ROI提取的前期处理——图像分割时,采用模糊聚类和边缘检测结合的方法实现较完整的提取目标。 3)在不能得到目标封闭轮廓的情况下,本文规定了不封闭轮廓的特征描述方法;拟订了针对这类图像的匹配策略。更多还原

【Abstract】 The image compression technology based on region of interest utilizes different compression to different area in an image; it can solve the contradiction between the image quality and the compression rate.ROI encoding is an important function of JPEG2000. ROI extraction is the first step to realize this function. Image segmentation can be used in this step. The paper completes extracting ROI in a group of image that always include a same object. The extracting method is: at first appoint a rectangle region that includes the object as ROI, extract the feature of ROI and record it. When search the ROI in other images, it uses the feature, which has been recorded to get the ROI. The main work of the paper include:1) The paper realizes ROI automatic extracting in a group of color image. The ROI extraction algorithms that have been published are mainly about grey image, they can’t be used in color image directly.2) Color image includes much information, it can’t get all pixel of object by a single method of image segmentation. The paper applies fuzz-clustering combine with edge detecting to solve this problem.3) In case that can’t get the closed edge of object, it defines not only the description method of object, but also the matching policy of this type of image.更多还原