Identification method of viewpoint type of video or image

Abstract

The invention discloses an identification method of a viewpoint type of a video or an image. The identification method of the viewpoint type of the video or the image comprises the following steps: obtaining a set histogram dependency threshold and an area similarity characteristic value threshold through training in advance; extracting areas needed for similarity detection according to an image arrangement rule for the inputted video or image; carrying out extraction of the histogram similarity characteristic value of the areas for similarity detection, then comparing the similarity characteristic value with the threshold to judge whether a result is similar, if the result is not similar, dividing the two areas into a plurality of blocks correspondingly, carrying out macro block likelihood extraction between the blocks, averaging all the macro block likelihood rate of the corresponding blocks of the two areas to be used as the similarity characteristic value, comparing the similarity characteristic value with the area similarity characteristic value threshold to judge whether a result is similar. According to the identification method of the viewpoint type of the video or the image, after a plurality of pairs of areas needed for similarity detection are extracted, the video source type can be judged. The identification method of the viewpoint type of the video or the image is based on the idea that the integrity similarity detection is primary and the local similarity detection is auxiliary, and is fast and accurate in identification.

Description

technical field [0001] The invention relates to the field of video processing and identification research, in particular to a video or image viewpoint type identification method. Background technique [0002] With the development of display technology and digital multimedia technology, people are more and more inclined to real visual experience. Stereoscopic display technology is regarded as the next generation display technology. And form a 3D model, and then feel that the object has a distance, like a real three-dimensional scene. At present, 3D displays can display 2, 8, 9 video sources of various viewpoints, and some 3D displays can even switch between 2D and 3D. The variety of videos greatly enriches the 3D world and rapidly promotes the application of 3D display technology. [0003] 2D flat-panel display is the mainstream of video now. For 3D displays, it is undoubtedly a wise choice to be compatible with 2D. Two viewpoints are left and right picture format video sou...

AI Technical Summary

Problems solved by technology

[0003] 2D flat-panel display is the mainstream of video now. For 3D displays, it is undoubtedly a wise choice to be compatible with 2D. Two viewpoints are left and right picture format video sources, including various forms of polarized light. Red and blue forms are currently the mainstream of various 3D displays. The video format can be implemented relatively easily. Users only need to wear red and blue glasses to see the 3D effect, but this kind of video needs to wear glasses, and the experience effect is relatively poor, and it can only be seen from a single angle. Relatively speaking, multi-view technology with formats such as 8 viewpoints and 9 viewpoints is the main direction of future development, which can show better effects on naked-eye 3D. For future stereoscopic display devices, if they are fully compatible with 2D and two Viewpoint, multi-viewpoint, its application range will be greatly expanded

[0004] In this regard, the current video format has not yet formed a unified standard, so the automatic identification of the video format has become a difficult problem. The existing methods include identification by adding additional identification codes, etc. However, due to the different formats of video layouts from different viewpoints At the same time, the internal rules of different companies are not the same, so the efficiency is low, the operation is complicated, and a unified standard cannot be formed

Images