The Uniform Local Binary Pattern(ULBP)and its many variants have shown effectiveness in texture classification.However,most of these ULBP methods focus on the uniform pattern features between the coding center pixel and its adjacent pixels.Therefore,they cannot capture the image features between pixels that occur outside the non-uniform mode,and they ignore the interaction between pixels at different scales.In response to these issues,the Extended Uniform Local Binary Pattern(EULBP)is proposed.EULBP counts the non-uniform patterns between adjacent pixels of the image,and gradually encodes the non-uniform patterns between adjacent sampling points.Secondly,extract image features at different scales for fusion.Finally,the similarity of feature vectors is calculated using the cross distance of histograms to obtain recognition rates in different datasets.The effectiveness of the algorithm has been verified through experiments.
关键词
扩展统一局部二值模式/图像特征提取/特征融合
Key words
Extended Uniform Local Binary Patterns/image feature extraction/feature fusion
维普
万方数据