A new solution in recovering video to video by hash method

Every day, with the advancement of visual equipment, large amounts of data generated in the form of films and videos and uploaded to the Internet. Extracting the desired video from an image has become an important challenge, and so, many specialists and experts have tried to provide various solutions, each with its strengths and weaknesses. A content-based video retrieval system consists of three basic steps: key frame extraction, important features extraction and similarity comparison. Hashing is one of the methods used for retrieving data, which is mostly used to retrieve images. In this paper, we propose a new framework using the hashing method to solve the video retrieval problem, which takes advantage of a multidimensional (3D) CNN to obtain the spatial and temporal features of the video. In the proposed method, the features extracted from each key form are transferred using the Hashing function to a binary space by the pre-trained network to receive the compressed binary codes of the video. We have done some test on the two video datasets THUMOS'14 and UCF-101, and the results show that the in proposed method than the existing methods, the value of mAP in the THUMOS'14 dataset increased 0.61% and 0.62%. in UCF-101 dataset.