6533b82afe1ef96bd128ca43

RESEARCH PRODUCT

Trademarks recognition based on local regions similarities

Fella HachoufAhmed ZeggariSebti Foufou

subject

Similarity (geometry)business.industryComputer scienceMathematics::History and OverviewComputingMethodologies_IMAGEPROCESSINGANDCOMPUTERVISIONCorner detectionPattern recognitionImage segmentationContent-based image retrievalEdge detectionComputingMethodologies_PATTERNRECOGNITIONComputer Science::Computer Vision and Pattern RecognitionPattern recognition (psychology)Computer visionArtificial intelligencebusinessCluster analysisImage retrieval

description

This paper deals with content based image retrieval. We propose a logo recognition algorithm based on local regions, where the trademark (or logo) image is segmented by the clustering of points of interest obtained by Harris corners detector. The minimum rectangle surrounding each cluster is detected forming the regions of interest. Global features such as Hu moments and histograms of each local region are combined to find similar logos in the database. Similarity is measured based on the integrated minimum average distance of the individual components. The results obtained demonstrate tolerance to logos distortions such as rotation, occlusion and noise.

yearjournalcountryeditionlanguage
2010-05-0110th International Conference on Information Science, Signal Processing and their Applications (ISSPA 2010)
https://doi.org/10.1109/isspa.2010.5605559