Editorial Image Retrieval Using Handcrafted and CNN Features

Citation

Companioni-Brito C, Elawady M, Yildirim S & Hardeberg JY (2018) Editorial Image Retrieval Using Handcrafted and CNN Features. In: Mansouri A, El Moataz A, Nouboud F & Mammass D (eds.) Image and Signal Processing. Lecture Notes in Computer Science, 10884. ICISP 2018: International Conference on Image and Signal Processing, Cherbourg, France, 02.07.2018-04.07.2018. Cham, Switzerland: Springer International Publishing, pp. 284-291. https://doi.org/10.1007/978-3-319-94211-7_31

Abstract
Textual keywords have been used in the early stages for image retrieval systems. Due to the huge increase of image content, an image is efficiently used instead according to the time computation. Deciding powerful feature representations are the important factors for the retrieval performance of a content-based image retrieval (CBIR) system. In this work, we present a combined feature representation based on handcrafted and deep approaches, to categorize editorial images into six classes (athletics, football, indoor, outdoor, portrait, ski). The experimental results show the superior performance of the combined features among different editorial classes.

Keywords
Image features; Similarity; CBIR; CNN; LBP; BoVW