Catálogo de publicaciones - libros

In this study, we present a method to extensively reduce the number of retrieved images and increase the retrieval performance for the person queries on the broadcast news videos. A multi-modal approach which integrates face and text information is proposed. A state-of-the-art face detection algorithm is improved using a skin color based method to eliminate the false alarms. This pruned set is clustered to group the similar faces and representative faces are selected from each cluster to be provided to the user. For six person queries of TRECVID2004, on the average, the retrieval rate is increased from 8% to around 50%, and the number of images that the user has to inspect are reduced from hundreds and thousands to tens.

Shape descriptors have been used frequently as features to characterize an image for classification and image retrieval tasks. For example, the patent office uses the similarity of shape to ensure that there are no infringements of copyrighted trademarks. This paper focuses on using machine learning and information retrieval techniques to classify an image into one of many classes based on shape. In particular, we compare Support Vector Machines, Naïve Bayes and relevance language models for classification. Our results indicate that, on the MPEG-7 database, the relevance model outperforms the machine learning techniques and is competitive with prior work on shape based retrieval. We also show how the relevance model approach may be used to perform shape retrieval using keywords. Experiments on the MPEG-7 database and a binary version of the COIL-100 database show good retrieval performance.

Evaluation of retrieval performance is a crucial problem in exhaustive web-based image retrieval. The challenge is to find reliable techniques for estimating the full set of matching answers. In this paper, we present an automatic image-gathering method based on keyword-based web search engines. User’s interaction is involved to build an initial reference image set and verify the result images. By filtering images based on the Earth Mover’s Distance on color feature and combining the results from several search engines, our approach achieves high accuracy in collecting images relevant to user’s information need. We have compared the result of our approach to the performance of three commercial web image search engines.

Given a collection of images and a set of image features, we can build what we have previously termed NN networks by representing images as vertices of the network and by establishing arcs between any two images if and only if one is most similar to the other for some weighted combination of features. An earlier analysis of its structural properties revealed that the networks exhibit small-world properties, that is a small distance between any two vertices and a high degree of local structure. This paper extends our analysis. In order to provide a theoretical explanation of its remarkable properties, we investigate explicitly how images belonging to the same semantic class are distributed across the network. Images of the same class correspond to subgraphs of the network. We propose and motivate three topological properties which we expect these subgraphs to possess and which can be thought of as measures of their compactness. Measurements of these properties on two collections indicate that these subgraphs tend indeed to be highly compact.

To detect the right position of liver objects in ultrasound image is a critical issue in medical image analysis and visualization. Most ultrasound image segmentation techniques focus on region growing and Active contours. These are semi-automatic segmenting systems because these approaches need a user to identify a seed point or to draw an initial contour. This paper proposes a novel automatic segmenting system to detect liver in ultrasound images. The peak-and-valley is adapted by scanning pixel along with the Hilbert curve. A “local adaptive threshold” procedure is proposed to further reduce noise from the images. After Otsu segmentation algorithm is applied to the images, a core area algorithm is employed to detect liver objects with the help of a feature knowledge base. The proposed method is compared with other techniques and the manual segmentation method. The results indicate the accuracy of the proposed system and these automatically segmented images contain less noise than the other methods. This system supports automated liver detection in ultrasound images.

This paper presents a new approach for building semantic image indexing and retrieval systems. Our approach is composed of four phases : (1) knowledge acquisition, (2) weakly-supervised learning, (3) indexing and (4) retrieval. Phase 1 is driven by a visual concept ontology which helps the expert to define low-level features useful to characterize object classes. Phase 2 uses acquired knowledge and image samples to learn the mapping between image data and visual concepts. Image indexing phase (phase 3) is fully automatic and produces semantic annotations of the images to index. The symbolic nature of querying enables user-friendly and fast retrieval (phase 4). We have applied our approach to the domain of transport vehicles (i.e. motorbikes, aircrafts, cars).

We analyze the special structure of the relevance feedback learning problem, focusing particularly on the effects of image selection by partial relevance on the clustering behavior of feedback examples. We propose a scheme, aspect-based relevance learning, which guarantees that feedback on feature values is accepted only once evidential support that the feedback was intended by the user is sufficiently strong. The scheme additionally allows for natural simulation of the relevance feedback process. By means of simulation we analyze retrieval performance, search regularity and sensitivity to feature errors.

The performance of a new content-free approach to image retrieval is demonstrated. Accumulated user feedback data that specify which images are (ir)relevant to each other and keywords obtained from a network game are recycled through collaborative filtering techniques to retrieve images without analyzing actual image pixels. Experimental results show the proposed method outperforms a conventional content-based approach using support vector machine. The result was achieved by the combination of feedback data and keywords. Applications of the proposed scheme in query-by-text image retrieval is also discussed.

In this paper, we propose a novel paired feature learning system for relevance feedback based image retrieval. To facilitate density estimation in our feature learning system, we employ an ID3-like balance tree quantization method to preserve most discriminative information. In addition, we map all training samples in the relevance feedback onto paired feature spaces to enhance the discrimination power of feature representation. Furthermore, we replace the traditional binary classifiers in the AdaBoost learning algorithm by Bayesian weak classifiers to improve its accuracy, thus producing stronger classifiers. Experimental results on content-based image retrieval show improvement of each step in the proposed learning system.