中文摘要：

该文在AdaBoost算法的基础上提出了一种图像局部区域相似度的学习架构,利用该架构训练图像局部特征来获得低维数、独特的特征描述子,以实现对图像局部区域高精度地匹配.所提学习架构通过学习图像局部区域相似性得到一组非线性弱学习器对图像局部特征进行描述;同时,在响应函数组合形式和弱学习器权重优化配置方面,针对浮点描述子和二值描述子分别提出了新的补丁相似性度量函数作为目标函数的核函数,提高了图像特征相似性匹配效果.该学习架构不会受限于任何预定义的图像特征信息采集模式,能产生基于灰度信息或方向梯度信息的特征描述子.实验结果表明采用这种学习架构获得的特征描述子,在所有对比描述子中图像局部匹配查准率是最好的.所提学习框架能有效地配置优化描述子弱学习器,能提高图像特征描述子对图像尺度和视角变化的鲁棒性.

英文摘要：

A learning framework of image patch similarity to learn low-dimensionality and highdiscriminative descriptor based on AdaBoost is proposed.In the framework,the representations of image patches are modeled by non-linear weak learners which are trained by AdaBoost algorithm.Meantime,in the aspect of the response function combination and the weak learners＇ weight optimization allocation,two new similarity functions for float point descriptor and binary descriptor respectively are proposed to use as kernel function of optimized object function to learn a similarity embedding for improving image feature matching effect.The proposed learning framework is more generalizing than others as it won＇t be restricted to any predefined featuresampling model and it could encompasses intensity and director-gradient information.The results show that the proposed feature descriptor outperforms overall comparing descriptors on recallprecision performance of image local area matching.The proposed learning framework is able to effectively optimize over the descriptor filter configuration leads to boost robustness to image scale and perspective changes.