Scene classification apparatus of video

A scene classification apparatus of video for classifying a sequence of shots into a slow scene, where the slow scene includes a plurality of continuous shot and is thus a larger unit than a shot, comprising:

• a shot segmentation device to segment the video into respective shots based on cut points;
• an extractor for extracting from the respective segmented shots a shot similar to a current target shot from shots after a shot before the current target shot only by a predetermined interval;
• and a slow scene detector for classifying the target shot into a slow scene of the shot similar to the current target shot based on motion intensity of the current target shot and the shot similar to the current target shot, wherein each scene includes a plurality of cut points;
• wherein the slow scene detector classifies the target shot into the slow scene of the shot similar to the current target shot when the motion intensity of the shot similar to the current target shot is stronger than the motion intensity of the current target shot.

A scene classification apparatus of video for classifying a sequence of shots into a scene in which a camera operation has been performed, where the scene in which the camera operation has been performed includes a plurality of continuous shots and is thus a larger unit than a shot, comprising:

• a shot segmentation device to segment the video into respective shots based on cut points;
• a detector for detecting a histogram relating to motion directions of the respective segmented shots;
• and a detector for detecting the scene in which the camera operation has been performed based on the histogram of motion direction, wherein each scene includes a plurality of cut points;
• detector for detecting spatial distribution of motion of each shot;
• and a panning scene detector for detecting whether the respective shots are a panning scene based on the histogram of motion direction and the spatial distribution of motion.

A scene classification apparatus of video for segmenting video into shots based on cut points and classifying each scene composed of one or more continuous shots based on a content of the scene, comprising:

• a detector for detecting a highlight scene;
• extracting and combining means for extracting and combining a plurality of highlight scenes;
• and inserting means for inserting a video transition effect into a combined portion of the respective highlight scenes, the inserting means including a dynamic/static scene detector to detect whether a highlight scene is a dynamic scene with much motion or a static scene with little motion, wherein the inserting means makes a type of the video transition effect to be inserted different according to whether the highlight scenes to be combined are the dynamic scene or the static scene, wherein each scene includes a plurality of cut points;
• the highlight scene is the dynamic scene, the video transition effect with small change in an image mixing ratio is inserted therein, and when the highlight scene is the static scene, the video transition effect with large change in the image mixing ratio is inserted therein.

A scene classification method of video for classifying a sequence of shots into a dynamic scene with much motion or a static scene with little motion, where the dynamic scene and the static scene respectively include a plurality of continuous shots and are thus a larger unit than a shot, comprising:

• segmenting the video into respective shots using a shot segmentation device that segments the video based on cut points;
• calculating shot density DS of the video per a time unit from the respective segmented shots;
• calculating motion intensity per unit region on an image using a value of motion vectors of the respective segmented shots;
• and
• classifying the sequence of shots into the dynamic scene with much motion, the static scene with little motion or an other scene except the dynamic scene and static scene based on the shot density and the motion intensity of the respective segmented shots, wherein each scene includes a plurality of cut points,
• wherein the classifying classifies a sequence of shots whose shot density is larger than a first reference density and whose motion intensity is stronger than a first reference intensity into the dynamic scene, and classifies a sequence of shots whose shot density is smaller than a second reference density and whose motion intensity is weaker than a second reference intensity into the static scene, and
• wherein the classifying classifies a sequence of shots whose shot density is not larger than the first reference density or whose motion intensity is not stronger than the first reference intensity, and a sequence of shots whose shot density is not smaller than the second reference density or whose motion intensity is not weaker than the second reference intensity as neither the dynamic scene nor the static scene.