Neighborhood ternary co-occurrence for natural and texture image retrieval

In the today’s digital scenario, everything is turning into digital. It is due to the ease in access and storing of digital media. But it also creates complexity in searching particular image from the huge repository. In the solution text based image retrieval was proposed but it is time consuming and prone to human errors. Content based image retrieval (CBIR) solves the problem by automatically representing images by the image contents. Color, texture, shape etc. are used to extract image features and these features are used to match images. The objective of this research papers is to introduce a new content based feature which is suitable for different types of images. In this paper, a novel pattern feature neighborhood ternary co-occurrence pattern (NTCoP) is proposed. It extracts directional information in eight major angles. Second order derivative highlights the edges. Ternary pattern of those edges are used to compute statistics using co-occurrence. Results on two very popular datasets, one each from natural scenes and textures prove the robustness of the proposed feature as compared to the many state-of-the-art methods. The main highlight of the proposed feature is to capture edge information along with local textural statistics. This simple yet important information helps to represent variety of images efficiently and perform image retrieval on diversified datasets. Standard metrics are used to compare the performance and proposed method has shown better precision and recall. © 2023, The Author(s), under exclusive licence to Bharati Vidyapeeth's Institute of Computer Applications and Management.