A ROI image retrieval method based on CVAAO

Semantic content-based image retrieval: A comprehensive study

“…However, these methods employ a fixed size of regions and do not consider multiple ROIs for the similarity matching between different ROIs of different images. Multiple ROIs [150][151][152] are employed to provide relative locations of multiple ROIs, and it considers the other blocks that have different spatial locations from ROIs in the image. Lee and Nang [150] have used the MPEG color dominant as feature extracted from image blocks, and selected the blocks having a higher overlapping area to overlap them with user's identified regions.…”

ROI image retrieval based on multiple features of mean shift and expectation–maximisation

“…There are several evaluation measures such as rank, precision, recall, precision/recall graph, etc. presented in [27] and a kind of ranking measure, ANMRR (Average Normalized Modified Retrieval Rank) that is defined from the MPEG-7 research group [45,4]. In order to objectively evaluate the proposed method, in our experiments we use the two indicator groups.…”

“…So far, the region of interest, i.e. (ROI)-based query, among these is the most common method in research prototypes and commercial products [1,4,9,14,15,16,23,28]. Typically a query image contains not only the ROIs but also other irrelevant areas.…”

“…Chan et al [4] proposed an image retrieval method based on colour variance among adjacent objects (CVAAO), which determines the location and size of the ROI in the current image through features such as shape, area, and position of the largest object, and then cuts off the ROI from the image. The method is also compared to some conventional methods such as CCH, FCH, and SamMatch and confirms it outperforms these conventional methods.…”

Color Image Retrieval Based on Full Range AutoRegressive Model with Low-Level Features

“…Mainly, the key issue to develop a successful retrieval system relies on identifying and choosing the right features that represent the images as strong as possible. Feature representation of the images may include color [1,2,3,4,5,6,9], texture ( [3,7,10,11,13] and shape [1,6,7,8] information.…”