Content based image retrieval (CBIR) systems provide a faster way to retrieve images by representing them in terms of their visual contents. In this paper, a novel texture feature, directional local ternary co-occurrence pattern (DLTCoP) is proposed for CBIR. First and second order derivatives of the image are extracted through directional filter masks to capture coarse and fine details of the image in four directions. Thereafter, changes in first and second order filter responses are analyzed simultaneously and co-occurrence is computed based on their inter-relations. The information captured by DLTCoP is further enriched by computing histograms for the gray-scale image and the color information is represented as color histograms. The proposed scheme provides a consolidated feature capable of distinguishing between different images. Experiments are conducted on five benchmark data sets, Corel 1000, Corel 5k, Corel 10k, INRIA Holidays and Salsburg Texture. Significant improvement in average precision and recall is obtained with respect to the existing state-of-the-art features.