Complete Local Spatial Relationship Pattern: An Efficient Texture Feature Descriptor for Image Retrieval

• Chelladurai Callins Christiyana

  Department of Computer Science and Engineering, SRM Madurai College for Engineering

• Murugesan Poomani Punitha

  Department of Information Technology, Sethu Institute of Technology

The texture is an exemplary feature of an image, and is exploited to delineate the images for visual content-based image retrieval (CBIR) approaches. Local pattern based texture features are very promising for representing image features. The local patterns are computationally efficient and discriminative for retrieving similar images. This article proposes a new local pattern called the Complete Local Spatial Relationship Pattern (CLSRP) to represent the texture for image retrieval. CLSRP is different from other local patterns, as it partitions the neighbouring pixels in the local (regional) neighbourhood according to the spatial distance from the center pixel of the neighbourhood and enforces the weights based on the spatial distance while encoding the patterns. The magnitude and sign factors of the patterns from the neighbourhood relationships and the center pixel binary mapping concerning the global thresholding are considered to characterize the CLSRP in the local neighbourhood and make the CLSRP pattern complete. The impact of CLSRP in CBIR is experimentally demonstrated in two standard databases, the ORL and UIUC databases, and real time Covid19 chest Xray database. The retrieval efficiency of CLSRP is presented by precision and recall metrics. The retrieval efficiency comparison demonstrates the notable improvements in the CLSRP over other cutting-edge local patterns.

• PDF

Copyright (c) 2026 Journal of Control Engineering and Applied Informatics

This work is licensed under a Creative Commons Attribution 4.0 International License.