Expert content-based image retrieval system using robust local patterns

• A new robust LBP (RLBP) algorithm for CBIR is proposed.
• RLBP operator is employing robust S_LBP and robust M_LBP.
• The combination of Gabor transform and RLBP operator has also been presented.
• Performance is tested under different lighting and noise conditions.
• Proposed method shows a significant improvement as compared to other methods.

A new image indexing and retrieval algorithm for content based image retrieval is proposed in this paper. The local region of the image is represented by making the use of local difference operator (LDO), separating it into two components i.e. sign and magnitude. The sign LBP operator (S_LBP) is a generalized LBP operator. The magnitude LBP (M_LBP) operator is calculated using the magnitude of LDO. A robust LBP (RLBP) operator is presented employing robust S_LBP and robust M_LBP. Further, the combination of Gabor transform and RLBP operator has also been presented. The robustness is established by conducting four experiments on different image database i.e. Corel 1000 (DB1), Brodatz texture database (DB2) and MIT VisTex database (DB3) under different lighting (illumination) and noise conditions. Investigations reveal a promising achievement of the technique presented when compared to S_LBP and other existing transform domain techniques in terms of their evaluation measures.

Figure optionsDownload high-quality image (142 K)Download as PowerPoint slide