基于多核学习的多特征融合图像分类研究

图像分类任务是计算机视觉中的一个重要研究方向。组合多种特征在一定程度上能够使得图像分类准确度得到提高。然而，如何组合多种图像特征是一个悬而未决的难题。提出了一种基于多类多核学习的多特征融合算法，并应用到图像分类任务。算法在有效地利用多核学习自动选取对当前任务有价值特征的优势的同时，避免了在多核学习中将多类问题分解为多个二分问题。在图像特征表示方面，使用字典自学习方法。实验结果表明，提出的算法能够有效地提高图像分类的准确度。

Multiple feature fusion via multiple kernel learning for image classification

Image classification is one of the most fundamental problems of computer vision. Fusing many features can improve the accuracy for image classification. However, how to fuse these different features is still an open question. A multiple feature fusion method based on multiclass multiple kernel learning is proposed and applied to image classification problem. The proposed method can effectively avoid dividing the multiple class problem into many binary classification, and obtain the classifier directly. An adaptive dictionary learning algorithm for sparse coding is also introduced. Experimental results show that the proposed method can improve the accuracy for image classification significantly.