Research on Key Technology for Game VIdeo Retrieval

The paper puts forward that conducting lens boundary detecting by putting image-based time-space module features into application. The method firstly constructs a large number of time-space modules, namely, it detects lens boundary through calculating special vector of pixel in each time-space module, transferring the vector into a 3x3 matrix, and judging whether the matrix has answers or not. Identical subject will be classified into different sub modules due to subjective factors when structuring time-space modules, which ultimately will result in false and miss-judgment of lens boundary detecting. Putting forward the image-based method of time-space combination, the paper determines whether merge two sub modules by comparing similarity between each sub section in initial frame and its adjacent area. It can be proved through ultimate experiment that the approach employed by the paper is capable of producing good results in promoting precision of lens boundary detecting for game technology.