摘要
综述了字典学习算法的主要研究方向之一,即以图像分类为目标的稀疏表示字典学习算法。从空间变换法和类别指示法两个角度,分析各种算法的优缺点,并对相应的实验结果进行比较。总结了利用这类算法进行图像分类时所面临的其他一些关键问题,如模式识别中的旋转不变性和计算速度等。依据目前已有的技术和应用需求,探寻该领域未来的研究方向。
One of the main research directions of dictionary training algorithm,which is also called sparse representation dictionary learning algorithm for image classification,is summarized.The merits and demerits of various algorithms are analyzed in the aspects of spatial alternation and category indication methods.The relevant experiment results are compared.Some other key problems existing in image classification by such algorithms are summed up,such as rotatioal invariance and computation speed in pattern recognition.Based on existing technology and application requirements,some future research directions in this field were explored.
出处
《现代电子技术》
2013年第2期22-25,28,共5页
Modern Electronics Technique
基金
中央高校基本科研业务费专项资金资助项目(CHD2012JC012)
陕西省社会发展科技攻关项目(2010K11-02-11)
陕西省教育厅科学研究计划资助项目(11JK0994)
关键词
图像分类
稀疏表示
字典训练
原子
image classification
sparse representation
dictionary training
atom
