透過色調分析於影片分類之研究

劉佳樺 ； 陸清達 ； 沈俊宏 ； 王玲玲

《TANET2017 臺灣網際網路研討會》 2017年 (2017/10) Pp. 1465-1470

https://doi.org/10.6728/TANET.201710.0257

摘要

影片數目眾多而龐雜，電影種類也極為混雜，對於一部未知的影片，觀眾很難知道此影片的類型，造成觀眾想要選擇一部想要觀看類型的影片，必須花費很多時間觀看預告片或跳躍式播放該段電影，以便了解該影片是否屬於自己想要觀賞的類型。本研究的目的是要讓觀眾無須先行觀看預告片或影片內容就可知曉影片類型，便可以決定是否屬於自己想看的影片類型。在本論文中，我們提出透過色調分析，作為影片分類的依據；首先將影片以取樣方式擷取影格，再將影格中色調分類為紅色、綠色、藍色及黃色等四類，做為影片分類的特徵參數。我們根據影格色調的數量、連續變化性將影片分類成文藝愛情類、科幻類、動作冒險類、恐怖驚悚類、喜劇等五大類。經實驗結果發現，文藝愛情類及喜劇類以紅色及黃色系居多、科幻類及恐怖驚悚類綠色及藍色系居多、動作冒險類東洋片以黃色系及紅色系居多，西洋片則以藍色系及綠色系居多，這項研究結果可以應用於影片類型的自動分類。

關鍵字

影片分類 ； 色調分析 ； 影格擷取 ； 色調連續性 ； 多數決定機制

並列摘要

There is a great quantity of films over networks, such as the Youtube web site. The types of the films are various and complex. A viewer is difficult to know whether the type of a film is interesting to her/him before watching it, so the viewer has to spend a lot of time in watching the trailer or the film with skip mode to select an interested film. This study aims to let viewers know the type of the movie without having to watch the trailers or the film content first. Initially, the film is sampled to capture the frames. In turn, the majority of the hues are analyzed and classified as red, green, blue and yellow categories for each frame. The majority and the continuous change of the hues are analyzed and employed to classify the films into literary love, science fiction, action adventure, horror thriller, and comedy categories. Experimental results reveal that literary and artistic comedy films are red and yellow majority, science fiction and horror thriller classes are green and blue majority, action adventure Dongyang films are yellow and red majority, western films are blue and green majority. These results can be applied to the automatic classification of the film type.

並列關鍵字

film classification ； hue analysis ； frame capture ； variation continuity of hue ； majority decision