Locating Regions of Interest in CBIR with Multi-instance Learning Techniques

Abstract

Abstract. In content-based image retrieval (CBIR), the user usually poses several labelled images and then the system attempts to retrieve all the images relevant to the target concept defined by these labelled images. It may be helpful if the system can return relevant images where the regions of interest (ROI) are explicitly located. In this paper, this task is accomplished with the help of multi-instance learning techniques. In detail, this paper proposes the CkNN-ROI algorithm, which regards each image as a bag comprising many instances and picks from positive bag the instance that has great chance to meet the target concept to help locate ROI. Experiments show that the proposed algorithm can efficiently locate ROI in CBIR process.

A Convex Method for Locating Regions of Interest with Multi-instance Learning

“…We employ the image database that has been used by Zhou et al [29] in studying the ROI detection performance of multi-instance learning methods. This database consists of 500 COREL images from five image categories: castle, firework, mountain, sunset and waterfall.…”

“…Moreover, we further compare with three stateof-art methods on locating the ROIs, namely, Diverse Density (DD) [15], EM-DD [26] and CkNN-ROI [29]. For the MI-SVM, mi-SVM, MI-Kernel and KI-SVMs, the RBF kernel is used and the parameters are selected using cross-validation on the training sets.…”

“…Following [29], we evaluate the success rate, i.e., the ratio of the number of successes divided by the total number of relevant images. For each relevant image in the database, if the ROI returned by the algorithm is a real ROI, then it is counted as a success.…”

“…The second approach is the CkNN-ROI algorithm [29], which is a variant of CitationkNN [23]. This approach uses Citation-kNN to predict whether a bag is positive or not.…”

Score thresholding for accurate instance classification in multiple instance learning

“…In the past, it has been used for molecule activity prediction [2], image classification [3], computer aided diagnosis [4], visual object tacking [5] and document classification [6]. MIL research traditionally focuses on bag classification, however, more recently, several authors considered problems in which instance must be classified individually [4], [7], [8].…”