System and method for comparing an input digital video to digital videos using extracted and candidate video features

A system for managing digital videos, comprising:

• a feature extracting unit for extracting a feature of an input digital video;
• a feature searching unit having a plurality of features of a plurality of digital videos for searching for candidate features having similarity to the extracted feature;
• a feature comparing unit for comparing the extracted feature with each of the candidate features to recognize a matching digital video of the plurality of digital videos having at least one of the candidate features similar to the extracted feature and provide information on the matching digital video corresponding to the at least one similar candidate feature;
• and
• wherein the feature extracting unit includes:
• a video decoder for decoding the input digital video into RGB video frames;
• a spatiotemporal sampler for normalizing the decoded video frames to a uniform size;
• and
• a feature extractor for extracting, as the feature of the input digital video, temporal and spatial attributes of the normalized video frames, wherein the feature extractor includes:
• a temporal feature extractor for extracting a temporal feature from the normalized frames;
• a spatial feature extractor for extracting a temporal feature from the normalized frames;
• and
• a spatiotemporal feature extractor for extracting a spatiotemporal feature from the normalized frames, wherein the spatiotemporal feature extractor includes:
• a frame buffer for temporarily storing the normalized frames by a predetermined buffer size;
• a cubic generator for dividing the normalized frames into 3D-based cubics with predetermined width, height and length;
• a 3D transformer for per 3D transformation on the divided Cubics;
• and
• a linear/nonlinear transformer for performing linear transformation or nonlinear transformation on the 3D transformed cubics, wherein the predetermined width is determined by dividing a width of the stored normalized frames by a number of cubics in a horizontal direction, the predetermined height is determined b dividing a height of the stored normalized frames by a number of cubics in a vertical direction, and the predetermined length is equal to the predetermined frame buffer size.