Multi - Feature based Logo Identification and Retrieval from Document Images

Veershetty C

Abstract: This paper presents an effective method for logo identification and retrieval from document images, utilizing multiple features. Initially, logos are identified and segmented using a segmentation algorithm, and the resulting segmented logo images are stored in a library. Subsequently, a combination of Local Binary Patterns (LBP), Histogram of Oriented Gradients (HOG), and Texture features is extracted to enable discrimination between logos of different categories. The cosine distance metric is then applied to measure the similarity between a query logo and candidate logos, and relevance is determined by generating distance ranks. Through experimentation, a threshold of 98% is selected, and logos that match appropriately are retrieved. The proposed method achieves impressive performance, with an average precision rate of 92%, an average recall rate of 93.20%, and an average F - measure of 91.00%.

Keywords: Logo retrieval, Logo identification, Document image, Multi feature, Segmentation, Algorithm

How to Cite?: Veershetty C, "Multi - Feature based Logo Identification and Retrieval from Document Images", Volume 12 Issue 8, August 2023, International Journal of Science and Research (IJSR), Pages: 1722-1728, https://www.ijsr.net/getabstract.php?paperid=SR23814083658, DOI: https://dx.doi.org/10.21275/SR23814083658

Download Citation: APA | MLA | BibTeX | EndNote | RefMan