The Complete Local Spatial Central Derivative Binary Pattern for Ultrasound Kidney Images Retrieval
The Content Based Image Retrieval (CBIR) is an active research domain in medical applications. The feature extraction process is the vital procedure in CBIR. This work proposes a new feature extraction procedure named as Complete Local Spatial Central Derivative Binary Pattern (CLSCDBP) for ultrasound kidney images retrieval. In a local 3X3 square region of an image, the new pattern considers the relationships among the surrounding neighbors about their neighbors at different spatial distances whereas the standard Local Binary Pattern reflects the relationships between the center pixel and the surrounding neighbors. Though the surrounding neighbor pixels relationship has been considered in the Local Mesh Peak Valley Edge Patterns (LMePVEP), the proposed feature is different by deriving the local pattern based on the encoding of central derivative of the surrounding neighbors of the center pixel. The neighbors of each surrounding pixel in different spatial distances are considered during central derivative computation. The proposed local pattern becomes complete by accompanying the global mean statistics into it. The performance of this new feature is examined in ultrasound kidney images retrieval system. The experimental results confirm that CLSCDBP achieves considerable step up in the retrieval of ultrasound kidney images than LMePVEP in terms of Retrieval Efficiency.
Content Based Image Retrieval (CBIR), Texture, Local Patterns, Local Binary Pattern (LBP), Local Mesh Peak Valley Edge Patterns (LMePVEP), Ultrasound kidney images, Retrieval efficiency