A feature-based algorithm for detecting and classifying scene breaks

Abstract

We describe a new approach to the detection and classication of scene breaks in video sequences. Our method can detect and classify a variety of scene breaks, including cuts, fades, dissolves and wipes, even in sequences involving signi cant motion. We detect the appearance of intensity edges that are distant from edges in the previous frame. A global motion computation is used to handle camera or object motion. The algorithm we propose withstands JPEG and MPEG artifacts, even at very high compression rates. Experimental evidence demonstrates that our method can detect and classify scene breaks that are di cult to detect with previous approaches. An initial implementation runs at approximately 2 frames p e r second on a Sun workstation.

Modeling of Echocardiogram Video Based on Views and States

“…We use Sobel operator due to its smoothing effect which is important for noisy echocardiogram video. Then for two consecutive frames, edge ratio is computed in terms of the number of new edge pixels entering the frame and the number of old edge pixels leaving the frame [12] [16]. Exit-ing pixels are identified by keeping pixels in the first frame but not the second, and the entering pixels are identified by keeping pixels in the second frame and not in the first.…”

“…To detect view boundary, we use traditional color histogram based comparison [11] and edge change ratio [12]. After detecting shot boundary, we apply a novel technique for automatic view classification of each shot which is based on the signal properties and their statistical variations for each view in echo video.…”

Detection and Classification of Shot Transitions

“…A lower threshold is used to detect small differences that occur for the duration of the gradual transition while a higher threshold is used in the detection of shot cuts and gradual transitions. Zabih et al [9] proposed another method to detect edit effects by checking the spatial distribution of exiting and entering edge pixels. Both of these techniques are designed to detect cuts, fades and dissolves, so they have been used for comparison with our shot transition detection algorithm.…”

“…In Section 3, the algorithm for shot cut detection is extended to detect fades and dissolves. The proposed method uses block tracking to differentiate between changes caused by gradual effects from those caused by object and camera motion and it has been designed to handle some of the shortcomings of previous methods [9,10]. Experimental results confirming the validity of the approach are presented and discussed in Section 4.…”

Video Shot Boundary Detection Using Independent Component Analysis

“…In [2], scene changes were detected by using pixel difference and luminance histograms based on DC-images in compressed domains. In [3], edge changes were used as a feature for shot detection. Shot detection techniques were reviewed in detail in [4], and a statistical detector based on motion feature was proposed.…”