基于颜色特征提取的图像搜索引擎研究

发布时间：2018-06-28 18:26

本文选题：图像检索 + 颜色特征　；参考：《重庆理工大学》2012年硕士论文

【摘要】：随着网络技术、多媒体技术和数码技术的飞速发展，网络中的图像资源日益丰富起来。为了满足广大网络用户对图像检索的要求，各种基于Web的图像搜索引擎如雨后春笋般冒了出来。通常，人们在搜索图像时，最关心的是搜索结果是否符合用户的检索要求。而检索结果的准确性，是由图像匹配算法的优劣来决定的。首先，从基于颜色特征的量化算法、描述方法以及匹配算法入手，针对现有的一些只考虑颜色总体比例而忽略颜色具体空间分布的图像匹配算法的不足，提出了一种新的匹配算法，即基于分块的不规则图形相似比较的图像匹配算法。具体算法的创新之处在于：（1）改进算法融入了颜色的空间信息，使得检索结果更为准确。常用的颜色特征表示方法，如颜色直方图法、颜色主色法等，这些算法都只考虑了图像中颜色的整体比例，而忽略了图像中各种颜色具体的空间分布信息。这就导致了两幅具有相同颜色直方图的图像，可能因为其各自颜色分布不同，而使得两幅图像内容相差很大。（2）改进算法具有自动分块的思想，它可以根据图像中物体的特征自动分块，，然后再取两幅图像中各块进行匹配，提高了检索的准确性。某些改进的图像匹配算法，如基于分块的颜色直方图法或基于分块的颜色主色法等，通过对图像分块而考虑了颜色的空间信息。但是该类方法都是采用固定分块的方式来处理图像，而且其分块的数目没有结合实际图像中物体的特征来确定。一旦设定一个具体的分块数目后，不同的图像可能因为设定的块数与图像内容有很大的差别而使得图像的分割造成误差。其次，改进算法与其他算法进行了性能比较，设计并对系统进行了实现，然后通过实际检索结果的对比，对改进算法的性能及实用性进行了比较。实验证明，改进算法对图像的旋转、平移、尺寸等变化不相关，具有很好的稳定性，能够准确的检索出用户需求的图片。
[Abstract]:With the rapid development of network technology, multimedia technology and digital technology, the image resources in the network are becoming more and more abundant. In order to meet the requirements of the vast number of web users for image retrieval, a variety of Web-based image search engines sprang up. Usually, when people search for images, they are most concerned about whether the search results meet the user's requirements. The accuracy of retrieval results is determined by the merits and demerits of image matching algorithm. First of all, starting with quantization algorithm based on color feature, description method and matching algorithm, aiming at the shortcomings of some existing image matching algorithms which only consider the proportion of color and ignore the specific spatial distribution of color. In this paper, a new matching algorithm is proposed, that is, an image matching algorithm based on block similarity comparison of irregular graphics. The innovations of the algorithm are as follows: (1) the improved algorithm incorporates color spatial information to make the retrieval results more accurate. The commonly used color feature representation methods, such as color histogram method, color master color method and so on, all of these algorithms only consider the overall proportion of the color in the image, but ignore the spatial distribution information of various colors in the image. As a result, two images with the same color histogram may differ greatly because of their different color distribution. (2) the improved algorithm has the idea of automatic partitioning. It can be automatically divided into blocks according to the features of the objects in the image, and then each block of the two images can be matched to improve the accuracy of the retrieval. Some improved image matching algorithms, such as the color histogram method based on block or the color master color method based on block, consider the spatial information of color by dividing the image into blocks. However, all of these methods deal with the image in a fixed block way, and the number of blocks is not determined by the characteristics of the object in the actual image. Once a specific number of blocks is set, different images may cause errors due to the difference between the number of blocks set and the content of the image. Secondly, the performance of the improved algorithm is compared with that of other algorithms, and the system is designed and implemented. Then, the performance and practicability of the improved algorithm are compared by comparing the actual retrieval results. Experiments show that the improved algorithm is not related to the image rotation, translation, size and other changes, has good stability, can accurately retrieve the user's needs of the picture.
【学位授予单位】：重庆理工大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TP391.41

【参考文献】