基于多特征融合重排序的图像检索技术研究与系统实现

发布时间：2018-04-27 20:23

本文选题：图像检索 + SURF　；参考：《西北大学》2016年硕士论文

【摘要】：基于内容的图像检索技术一直受到图像处理领域研究人员的广泛关注,在电子商务、军事侦查、视频监控和医学等领域有着广泛的应用。本文在研究基于内容的图像检索相关理论的基础上,重点围绕优化图像检索结果,设计并实现了一种新的基于多特征融合重排序的图像检索原型系统,论文的主要工作如下：1.提出了一种基于Krawtchouk色度分布矩的颜色特征检索方法,首先将图像从三维的RGB空间转换到二维对立色度空间,然后通过计算图像的Krawtchouk矩向量来描述颜色分布,进而实现图像检索。实验结果表明,该方法能有效地降低特征维数,优于HSV颜色直方图方法。2.实现了一种基于自反馈多特征融合的图像检索方法。该方法在实现基于Krawtchouk色度分布矩的颜色特征检索方法的基础上,结合Krawtchouk矩及Gabor小波,通过自反馈方法自动确定颜色、形状、纹理分量的相应权值,并进行多特征融合检索。实验验证了本文方法的有效性。3.设计并实现了一个基于多特征融合重排序的图像检索原型系统。该系统首先通过SURF-BoVW的图像检索方法进行初始检索,过滤掉图像库中相关性较差的图像,作为重排序检索的图像集；然后,利用基于自反馈多特征融合的图像检索方法进行重排序检索,输出优化的检索结果。实验结果表明,该系统在保证检索性能的同时,也具有较高的检索效率。
[Abstract]:Content-Based Image Retrieval (CBIR) technology has been widely concerned by researchers in the field of image processing, and has been widely used in e-commerce, military investigation, video surveillance and medicine. Based on the research of the theory of content-based image retrieval, a new image retrieval prototype system based on multi-feature fusion and reordering is designed and implemented around the optimization of image retrieval results. The main work of this paper is as follows: 1. In this paper, a color feature retrieval method based on Krawtchouk chrominance distribution moments is proposed. Firstly, the image is transformed from three-dimensional RGB space to two-dimensional opposite-chromaticity space, then the color distribution is described by calculating the Krawtchouk moment vector of the image, and then the image retrieval is realized. Experimental results show that this method can effectively reduce the feature dimension and is better than HSV color histogram method. 2. An image retrieval method based on self-feedback multi-feature fusion is implemented. On the basis of realizing the color feature retrieval method based on Krawtchouk chrominance distribution moment, combining Krawtchouk moment and Gabor wavelet, the corresponding weights of color, shape and texture components are automatically determined by self-feedback method, and multi-feature fusion retrieval is carried out. Experimental results show that the proposed method is effective. A prototype image retrieval system based on multi-feature fusion and reordering is designed and implemented. The system first carries out initial retrieval through SURF-BoVW 's image retrieval method, filters out the images with poor correlation in the image database, and serves as the image set for reordering and retrieval, and then, An image retrieval method based on self-feedback multi-feature fusion is used to reorder the retrieval and output the optimized retrieval results. The experimental results show that the system not only ensures the retrieval performance, but also has a high retrieval efficiency.
【学位授予单位】：西北大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.41

【相似文献】