ImageRover is a search by image content navigation tool for the world wide web. To gather images expediently, the image collection subsystem utilizes a distributed fleet of WWW robots running on different computers. The image robots gather information about the images they find, computing the appropriate image decompositions and indices, and store this extracted information in vector form for searches based on image content. At search time, users can iteratively guide the search through the selection of relevant examples. Search performance is made efficient through the use of an approximate, optimized k-d tree algorithm. The system employs a novel relevance feedback algorithm that selects the distance metrics appropriate for a particular query. Keywords: Image databases, query by image content, content-based retrieval, world wide web search engines. 1 Introduction For a while now there have been software "robots" roving the World Wide Web (WWW) collecting index information about th...

Object-based video representation, such as the one suggested by the MPEG-4 standard, offers a framework that is better suited for object-based video indexing and retrieval. In such a framework, the concept of a "key frame" is replaced by that of a "key video object plane". In this paper, we propose a method for key video object plane selection using the shape information in the MPEG-4 compressed domain. The shape of the video object is approximated using information on the shape coding modes in the MPEG-4 bitstream. Two popular shape distance measures, the Hamming and Hausdorff distance measures, are modified to measure the similarities between the approximated shapes of the video objects. Although they feature different computational and implementation complexity tradeoffs, the corresponding algorithms achieve essentially the same performance levels in selecting key video object planes that represent efficiently the salient content of the video objects. Key words: Key video ...

We propose an original approach for the characterization of video dynamic content with a view to supplying new functionalities for motion-based video indexing and retrieval with query by example. We have designed a statistical framework for motion content description without any prior motion segmentation, and for motion-based video classi cation and retrieval. Contrary to other proposed methods, we do not extract from a given video sequence a set of motion features but we identify a global probabilistic model, expressed as a temporal Gibbs random eld. This leads to de ne a e cient statistical motion-based similarity measure, relying on the computation of conditional likelihoods, to discriminate various motion contents. We have carried out experiments on a set of 100 video sequences, representative of various motion situations (temporal textures as re and crowd motions, sport videos, car sequences, low motion activity examples). We have obtained promising results both for the video classi cation step and for the retrieval process. 1

Scene is considered a good unit for indexing and retrieving data from large video databases. In this paper, we present a new content-based approach for detecting and classifying scene changes in video sequences. Our technique can detect and classify not only abrupt changes (i.e., hard cuts) but also gradual changes such as fades and dissolves. We compute background difference between frames, and use background tracking to handle various camera motions. Although our method processes significantly less data, it results in more semantically rich pieces (i.e., scenes). Our experiments on various types of videos indicate that the proposed technique is much less sensitive to the predefined threshold values, and is very effective in reducing the number of false hits. Our approach is particularly suitable for very large video databases because it is both space and time efficient. KEYWORDS: Video content analysis, scene detection, shot detection, video database management systems. 1. INTRODUC...

We present a novel methodology for the extraction of representative frames of a digital video sequence. The proposed method is called contentbased adaptive clustering(CBAC) which allows a user to focus on his interest in the video using these frames. It achieves this by allowing a user to select the preferred low-level content and the fraction of the frames he would like to extract from a video. In our algorithm, shot boundary detection is not needed. Video frames are treated as points in the multi-dimensional feature space corresponding to a low-level content such as color, motion, shape and texture. The changes of their distance are compared globally for extraction of representative frames. The frames of the video are dynamically clustered into two clusters according to their changes of distance. One cluster is designated for deletion and the other one is for retention. The algorithm converges to the result desired by the user by deleting some frames from the deletion cluster during each iteration. Based on our proposed CBAC method, we have developed a video player which has the functions of content-based browsing and content-based video summary. While it provides a flexible tool for video review, it can also be a sound basis for other work such as clustering of similar sequences and video retrieval. KEY WORDS: digital video, adaptive clustering, representative frame, content-based,

In this paper we propose to compare two region-based approaches to content-based video indexing and retrieval. Namely a comparison of a system using the Earth Mover’s Distance and a system using the Latent Semantic Indexing is provided. Region-based methods allow to keep the local information in a way that reflects the human perception of the content. Thus, they are very attractive to design efficient Content Based Video Retrieval systems. We presented a region based approach using Latent Semantic Indexing (LSI) in previous work. And now we compare performances of our system with a method using the Earth Mover’s Distance that have the property to keep the original features describing regions. This paper shows that LSA performs better on the task of object retrieval despite the quantification process implied. 1.

A constraint of existing content-based video data models is that each modeled semantic description must be associated with time intervals exactly within which it happens and semantics not related to any time interval are not considered. Consequently, users are provided with limited query capabilities. This paper is aimed at developing a novel model with two innovations: (1) Semantic contents not having related time information can be modeled as ones that do# (2) Not only the temporal feature of semantic descriptions, but also the temporal relationships among themselves are components of the model. The query system is by means of reasoning on those relationships.

Content-based Image Retrieval systems (CBIRS) start ourishing on the Web. Their performances are continuously improving and their base principles span a wide range of diversity. Content-based Video Retrieval systems (CBVRS) are less common and seem at a first glance to be a natural extension of CBIRS. In this document, we summarise advances made in the development of CBVRS and analyse their relationship to CBIRS. While doing so, we show that CBVRS are actually not so obvious extensions of CBIRS.

This study surveys current trends/methods in video retrieval. The major themes covered by the study include shot segmentation, key frame extraction, feature extraction, clustering, indexing and video retrieval-by similarity, probabilistic, transformational, refinement and relevance feedback. This work has done in an aim to assist the upcoming researchers in the field of video retrieval, to know about the techniques and methods available for video retrieval.

We present a complete and e#cient framework for video shot indexing and retrieval. Video shots are described by their key-frame, themselves described by their regions. Regionbased approaches su#er from the complexity of segmentation and comparison tasks. A compact region-based shot representation is usually obtained thanks to vector-quantization method. We thus introduce LSA to reduce the noise inherent to the segmentation and the quantization processes. Then to better capture the content of video shots, we propose two original methods. The first takes advantage of a multi-scale segmentation of frames while the second uses multiple frames to represent a shot. Both approaches require more computation time during the pre-processing but not for indexing and comparison tasks. Indeed the extra information is included in the original signatures of shots. Finally we introduce a relevance feedback loop to optimize the search and propose a new method to optimize the e#ect of LSA. In the experimental section, we make an evaluation of latent semantic analysis and proposed approaches on two problems, namely object retrieval and semantic content estimation.