基于内容的图像检索关键技术分析-计算机应用技术专业论文.docxVIP

摘 摘 要 基于内容的图像检索技术是当前的一个研究热点。本文针对这一课题，分别 在基于颜色、纹理、形状的检索方面进行研究，提出相应的解决算法。 基于颜色的图像检索方法，采用了DSQ算法对图像颜色进行初始量化，在此 基础上进行GLQ聚类，在进行图像匹配时，采用对称相似度算法。基于纹理的图 像检索方法，将彩色图像颜色由RGB空间转换到YIQ颜色空间上，对其三分量分 别进行树状小波分解，提取其纹理、颜色、纹理与颜色相关特征。在基于形状的 图像检索方法中，利用小波变换模极大值方法进行图像边缘提取，采用边缘像素 的梯度方向直方图来表示形状，它对图像亮度、平移、尺度具有不变性，在进行 图像匹配时，利用循环移位计算直方图的相关性克服了其对旋转敏感的缺陷。实 验结果证明，本文提出的方法使图像检索的性能得到很大提高。 关键词：基于内容的图像检索小波分析颜色量化纹理提取边缘检测 AbstractThe Abstract The technique of content—based image retrieval is a hot point of study nowadays． This paper deals with this subject in three basic features of images：cologtexture and shape，and brings forward an image retrieval algorithm for these three features separately． In color—based image retrieval，an improved algorithm is put forward．Color of the image is quantized by DSQ algorithm and clustered by GLQ algorithm．Symmetric resembility algorithm is employed in image matching．In texture-based image retrieval algorithm，the color of image is transformed from the RGB space to the YIQ space and tree-structured wavelet decomposition is applied to the three components separately．Then，features of texture，color and their relation are extracted and applied to image retrieval．In shape—based image retrieval algorithm，image edge is extracted by Wavelet transform based On maximum model algorithm，the shape is represented by histogram which shows the distribution of gradient directions at edge pixels．The histogram are invariant under lightness，translation and magnification of image，The relativity of two corresponding histogram which results from itertative rotation matching is used to avoid the sensitivity of the histogram of rotated image． Experimental results show that great improvement is achieved by the application of this algorithm to image retrieval． Key words：Content-based image retrieval Wavelet analysis Color quantization Texture extraction Edge detection 独创性(或创新性)声明本人声明所呈交的论文是我个人在导师的指导下进行的研究工作及所取得的 独创性(或创新性)声明 本人声明所呈交的论文是我个人在导师的指导下进行的研究工作及所取得的 研究成果。尽我所知，除了文中特别加以标注和致谢中所罗列的内容以外，论文 中不