如何對龐大的多媒體資料進行尋取是一重要的研究議題。本研究是以影像相似度為基礎,將影像相似度延伸應用至影片相似尋取,利用影像在空間關係上的相似度,進行影片相似度之計算。本研究之目的是針對一參考影片,查詢參考影片中與查詢影片最為相似之片段。我們從參考影片中,隨機挑選多張影像組成一段影片,並計算此段影片與查詢影片的相似度,但隨機組合之結果數量龐大,對其一一計算相似度不是有效率之方法,所以,本研究加入基因演算法,利用其尋求近似最佳解之能力進行影片相似尋取。 How to retrieve multimedia information is an important issue. Based on image similarity, our research extends to similarity retrieval for video. We measure video similarity according to spatial relation in images. Given a reference video, our objective is to retrieve the video segment in the reference video the most similar to query video. We select frames randomly from the reference video to compose them as a video segment and measure the similarity between it and the query video. But, the amount of random selection is very huge. It is not efficient to measure video similarity for each selection. Therefore, we propose the genetic algorithm to retrieve the video segment which is the most similar to query video in this paper.