Representation of image content based on RoI-BoW

Instance Search via Instance Level Segmentation and Feature Representation

“…There are basically two stages in visual content search system, namely feature representation [4,5,6,7,8,9,10,11,12,13,14,15,16] and fast retrieval [17,18,19,20]. In the whole process, feature representation plays the key role to the success of the system.…”

“…In the existing solutions, instance search has been mainly addressed by conventional approaches that are originally designed for image search [1,2], such as bag-of-visual words (BoVW) [4], RoI-BoVW [10], VLAD [5] and FV [9]. All these approaches are built upon image local features such as SIFT [21], Root-SIFT [22], SURF [23].…”

Image region annotation based on segmentation and semantic correlation analysis

“…No AIA algorithms could compete with the amazing image annotation capabilities of the human vision system [20]. Nonetheless a good image representation scheme with low-level image features will always improve the AIA performance [21]. In general two categories, global-based image content representation (GICR) and local-based image content representation (LICR), are utilised.…”

“…Now the BoW model has been widely used in image understanding and video retrieval [29][30][31][32][33][34], such as object recognition [29], image categorisation [22] and near duplicate detection [32]. BoW model can also be used to construct the model of syntactic and semantic relationship among the visual concepts, which makes it popular in constructing visual lexicons [21] and visual-semantic correlation analysis [33]. In this paper, we present a novel extension of the popular BoW model, which can represent the image region content more accurately by learning ensemble visual lexicons.…”