Abstract

We present an algorithm for video retrieval that fuses the decisions of multiple retrieval agents in both text and image modalities. While the normalization and combination of evidence is novel, this paper emphasizes the successful use of negative pseudo-relevance feedback to improve image retrieval performance. Although we have not solved all problems in video information retrieval, the results are encouraging, indicating that pseudo-relevance feedback shows great promise for multimedia retrieval with very varied and errorful data. 1 Video Retrieval from Mixed Text and Image Queries In this paper, we present an algorithm for the task of video retrieval. A query, consisting of a text description plus images or video is posed against a video collection, and relevant shots are to be retrieved. Our system accomplishes this by fusing the retrieval results of multiple retrieval agents. The overall system can be decomposed into several agents, including a text-oriented retrieval agent, which is responsible for finding the text in the speech transcripts [4] and Video OCR [3], a

Citations