最坏分离的联合分辨率判别分析

Worst-Separated Couple-Resolution Discriminant Analysis

现实中,常需辨识低分辨率(low-resolution,简称LR)图像(如监控系统所捕捉的人脸),但相比通常的高(high-resolution,简称HR)或超(super-resolution,简称SR)分辨率图像而言,其含有相对较少的判别信息,致使通常的子空间学习算法,如结合主成分分析(principal components analysis,简称PCA)的线性判别分析(linear discriminant analysis,简称LDA)难以获得理想的识别效果.为了缓和该问题,最近所提出的联合判别分析(如SDA)借助与低分辨率相配对的高分辨率图像辅助设计LR图像分类器.在SDA的实现中,其采用了类似LDA的平均散度定义,使SDA遗传了LDA在投影时难以使相对靠近的类充分分离的问题.为了克服该不足,提出了针对LR图像识别的最坏分离的联合分辨率判别分析(worst-separated couple-resolution discriminant analysis,简称WSCR),从而使:(1)LR和HR投影到同一低维子空间;(2)投影后的最小类间隔最大化.实验结果表明:与SDA相比,WSCR更适用于低分辨率的图像识别.

Low-resolution is an important issue when handling real world image recognition problems. The performance of traditional recognition algorithms, e.g. LDA/PCA, usually drops drastically due to the loss of discriminant information compared to those for high-resolution or super-resolution images. In order to solve this problem, many methods have been proposed in recent years based on coupled projections, i.e. learning two sets of different projections, one for high-resolution images and one for low-resolution images. For example, SDA （simultaneous discriminant analysis） obtains projections by maximizing the average between-class scatter while minimizing the average within-class scatters. Like LDA, SDA cannot separate projected classes, especially for those that are closer to each other. In this paper, a novel discriminant analysis method is proposed to achieve the optimal projections by maximizing the minimum distance between pair-wise classes. Experiments on several image datasets verify the efficiency of the presented methods.