采用多级区域局部二值模式(MB-LBP)对手势图像进行MB-LBP处理,获得手势多级区域复合LBP直方图向量,然后将多个复合LBP直方图向量进行特征匹配,最后算得手势的识别率。在Massey University手势数据库的实验结果表明,该方法与传统LBP方法相比,鲁棒性较好,识别率更高。
A multi-level regional Local Binary Pattern (MB-LBP) model is built to process MB-LBP gesture image data for obtaining the composite LBP histogram vectors of the multi-level gesture image. The vectors are matched with features to get the identification rate. Experimental results in Massey University gesture database show that with other one. the method is with higher identification rate, compared