| With the development of computer technology and multimedia technology, the Internet the rapid growth of video data. In the face of massive video data. In the face of massive video data, how to do video storage, organization, management and analysis, has become the main direction in the field of video. Develop and design an efficient video retrieval system is a meticulous and large workload, work involving knowledge broad and complex operation. A lot of video streaming data quantity, video scene, makes the same features presented in the video image texture feature is differ, may also be the same scenario there was some difference of image edge, these are likely to result in video retrieval precision rate and recall rate have deviation, bring difficulties to video retrieval.This article through to the video frame image to extract and analyze several characteristics, constructing the characteristic database, realizing fast matching of video image, mainly implements the following:(1) To study the current development of content-based video retrieval technology at home and abroad, summarize and analyze the key technology of content-based video retrieval.(2) Studied the video stream, the basic composing of detailed analysis of the image, this paper expounds the basic image processing algorithm in detail. By extracting the color of the image autocorrelation vector-valued, frame RGB channels, Gabor filter characteristic values for calculating grayscale average and variance and mean and variance of wavelet transform characteristic value, as the concrete basis of image retrieval.(3) Through experimental verification show that the proposed algorithm can rapidly retrieve the output in the system, and the output high accuracy, efficiency is higher, so to some extent, it has certain feasibility.The proposed feature extraction method, the acquisition of image library, based on pattern matching Manhattan, introduces the retrieval method of more detailed, more satisfactory effect through tests. |
PDF Full Text Request