This paper presents methods for video browsing and extracting information from video by measuring brightness data. Cut breaks are detected by measuring areas in which inter-frame difference occurs. We propose a video browsing tool (VBToo1) that creates a compressed video containing all important scenes. We utilize two factors related to pixel brightness to allow the automatic generation of the compressed video. . INTRODUCTIONThere have been many attempts at using video in computer and communication fields.2 However, there has been little improvement in the handling of video itself. Video cameras are ubiquitous but the information recorded on the tape is useless until someone looks at it and indicates the tapes's contents. Video information is difficult to handle because no machine can automatically "recognize" scenes with any degree of accuracy. Moreover, there are far more people with cameras producing video than the people capable of editing that video. We think it is necessary to convert video to make it easier to handle.The first stage of our study was an attempt to supply additional information for easy handling. This paper presents a method for extracting information relating to video content This information can then be used for video browsing. In section 2, we discuss the extraction method focusing on cut detection. In section 3, we propose a useful video browsing method as an example of how to utilize the information. The result of experiments are presented in section 4. EXTRACTION OF VIDEO CONTENT Extracting information from videoIt is useful to get information relative to video contents when you see, search, or edit videos. For example, knowing the location of cut breaks and converting video images into smaller units makes video easier to handle. For these reasons, we have to find characteristic values for cut breaks. Besides cut location, there is a lot of other useful information for handling. If the amount of motion and camera operation in a video are known, for instance contents can be recognized in a short time. The location of a slowmotion scene in an action movie is often indicates important or climactical scenes.To know this information, we use brightness data, because the brightness component is an essential value of images and it is well reflected in the video contents. We introduce two brightness values to extract different video content: frame-base histogram difference (FHD), and a pixel-base inter-frame difference histogram (IDH). (Fig. I) FHBrightness distribution is related to the image, and if images change. brightness distribution changes. We refer to the brightness histogram within a frame as FH(frame-base histogram). We can obtain information about moving images from the FH pattern. For example, at the beginning of cuts, FH changes discontinuously at the frame. (see Fig. 2 (a)) 980 / SPIE Vol. 1606 Visual Communications arid Image Processing '91: Image Processing o.9194ij7437/97/$4yJ Downloaded From: http://proceedings.spiedigitallibrary.org/ on 06/02/2015 Terms of Use: http://spied...

Abstract. This paper discusses a video cut detection method. Cut detection is an important technique for making videos easier to handle. First, this paper analyzes the distribution of the image difference V to clarify the characteristics that make V suitable for cut detection. We propose a cut detection method that uses a projection (an isolated sharp peak) detecting filter. A motion sensitive V is used to stabilize V projections at cuts, and cuts are detected more reliably with this filter. The method can achieve high detection rates without increasing the rate of misdetection. Experimental results confirm the effectiveness of the filter.

No abstract

No abstract

This paper proposes the application of tree-structured clustering to various noise samples or noisy speech in the framework of piecewise-linear transformation (PLT)-based noise adaptation. According to the clustering results, a noisy speech HMM is made for each node of the tree structure. Based on the likelihood maximization criterion, the HMM that best matches the input speech is selected by tracing the tree from top to bottom, and the selected HMM is further adapted by linear transformation. The proposed method is evaluated by applying it to a Japanese dialogue recognition system. The results confirm that the proposed method is effective in recognizing noise-added speech under various noise conditions.