项目名称: 退化四元小波关键技术研究
项目编号: No.61202319
项目类型: 青年科学基金项目
立项/批准年度: 2013
项目学科: 计算机科学学科
项目作者: 盖杉
作者单位: 南昌航空大学
项目金额: 23万元
中文摘要: 四元小波变换是一种新兴的图像多分辨率分析方法,也是一种有效的彩色图像描述手段。目前,四元小波变换的研究已经取得许多很好的成果尤其在彩色图像识别领域。然而,由于四元小波乘法规则不具有交换性,因此对于识别问题遇到了计算复杂度高与识别率难于提高的瓶颈。本项目通过对四元小波变换理论的认知,申请人采用创新的智能计算方法与改进的快速实现算法研究四元小波理论的延伸与拓展问题,主要研究内容包括(1)退化四元小波变换理论(退化四元傅立叶,卷积,相关性的数学表达与快速实现算法);(2)基于退化四元小波变换的多分辨率分析框架,研究和提出彩色图像在彩色空间中的表示方法;(3)针对低识别率的彩色图像识别系统,研究能提高识别性能的新的彩色图像识别系统,利用纹理图像分类试验评价模型的有效性。项目的预期研究成果将丰富和发展图像多分辨率分析的认知研究,突破目前一些传统彩色模式识别模型遇到的计算复杂度高与识别率难提高的瓶颈。
中文关键词: 退化四元小波;图像理解;稀疏表示;多分辨率分析;特征提取
英文摘要: Quaternion wavelet transform is new method of multi-resolution analysis for images. It is also an effective way for description of color images. The quaternion wavelet transform has obtained many better results in the domain of color image recognition. As the multiplication rule of the quaternion wavelet transform is not commutative, the recognition problem has high computation complexity and low recognition rate. The applicant made study on the extension of quaternion wavelet transform by new intelligent computation method and fast improved implementation algorithm. The main research content as follows: (1). The theory of reduced quaternion wavelet transform (the mathematic description of reduced fourier transform, convolution and correlation and implementation algorithms); (2) the presentation of color images in the color space based on frame of multi-resolution analysis; (3) According to the low recognition rate of color image recognition system, the new color image recognition system with high recognition rate is proposed. The effectiveness of the model is verified by the color image texture classification. Anticipated research results of the project will enrich and development of image multi-resolution analysis of cognitive research. Break the current conventional color pattern recognition model encounters
英文关键词: reduced quaternion wavelet;image understanding;sparse representation;multi-resolution analysis;feature extraction