查看文章

Image feature extraction can be effectively realized using local binary pattern (LBP). This paper proposes an image retrieval method using directional image features and an adaptive threshold. There are two steps: a. encoding with the standard deviation of the neighbors’ intensity values in the 3 × 3 pattern for threshold in center pixel. The threshold is adaptively updated concerning the intensity of the neighbor pixels. b. Estimation of the local directional pattern using the changes in the intensity of the neighborhood pixels in different directions along with binary encoding of neighboring pixels. These two features are combined and used for image retrieval. The proposed method is evaluated on Corel-1k data set and compared with other standard local feature representation methods.