Content Based Image Retrieval Using Local Directional Pattern (LDP) Image Descriptor

We present an approach for image retrieval using Local Directional Pattern (LDP) and efficient on-line learning. This paper presents an image retrieval using novel local feature descriptor, the Local Directional Pattern (LDP), for describing local image feature. A LDP feature is obtained by computing the edge response values in all eight directions at each pixel position and generating a code from the relative strength magnitude. Each bit of code sequence is determined by considering a local neighborhood hence becomes robust in noisy situation. A rotation invariant LDP code is also introduced which uses the direction of the most prominent edge response. Finally an image descriptor is formed to describe the image (or image region) by accumulating the occurrence of LDP feature over the whole input image (or image region). Experimental results on the UW database show that LDP impressively outperforms the other commonly used dense descriptors (e.g., Gabor-wavelet and LBP).

Keywords: Grey-Level Co-occurrence matrix(GLCM),Local Binary Pattern (LBP),k-NN,Rotation Invariant LDP.