Sumarização de vídeos de futebol com baixo custo computacional utilizando ritmo visual : descritores de cor e movimento de câmara

AUTOR(ES)
DATA DE PUBLICAÇÃO

2006

RESUMO

The growth of multimedia contents production has created the necessity of new techniques for efficient storage and retrieval of these contents. The area of Content Based Image Retrieval is a research field whose goal is to develop Computer Vision based methods for indexing and retrieval of images and videos. Visual features such as color, shape, texture and motion are used as descriptors computed directly from an image or video in CBIR applications. This descriptors are applied in different tasks such as shot transition detection, shot classification and objects recognition. The visual rhythm is a spatio-temporal sampled representation of video data providing compact information while preserving several types of video events. We exploit these properties in the current work to propose two new low level descriptors for the analysis of soccer videos computed directly from the visual rhythm. The descriptors are related to dominant color and camera motion estimation. The new descriptors are applied in different tasks aiming the analysis of soccer videos, such as shot transition detections, shots classification and attack direction estimation. We also present a simple automated soccer summary application to illustrate the use of the new descriptors. Keywords: low level video features, dominant color, motion estimation, soccer video analysis, mathematical morphology

ASSUNTO(S)

informÁtica - dissertaÇÕes sistemas de informacao

ACESSO AO ARTIGO

Documentos Relacionados