Efficient and Robust Detection of Duplicate Videos in a Large Database

Abstract

We present an efficient and accurate method for duplicate video detection in a large database using video fingerprints. We have empirically chosen the Color Layout Descriptor, a compact and robust frame-based descriptor, to create fingerprints which are further encoded by vector quantization. We propose a new non-metric distance measure to find the similarity between the query and a database video fingerprint and experimentally show its superior performance over other distance measures for accurate duplicate detection. Efficient search can not be performed for high dimensional data using a non-metric distance measure with existing indexing techniques. Therefore, we develop novel search algorithms based on pre-computed distances and new dataset pruning techniques yielding practical retrieval times. We perform experiments with a database of 38000 videos, worth 1600 hours of content.For individual queries with an average duration of 60 sec (about 50% of the average database video length), the duplicate video is retrieved in 0.032 sec, on Intel Xeon with CPU 2.33GHz, with a very high accuracy of 97.5%.

Effective and Efficient Image Copy Detection with Resistance to Arbitrary Rotation

“…The computation time of copy detection usually consists of offline processing time and online processing time [29]. Therefore, we analyze the time complexities of offline processing for test images from an image database and online processing for query images.…”

The MPEG-7 Video Signature Tools for Content Identification

“…In turn, such features are usually keypoint-based [11], [12], [13], [14], block-based [2], [15], [16] or global [17], [18]. For keypointbased approaches, SIFT features [14], [19], [20] or local descriptors around Harris interest points [12], [13], [21], [22] are a common design choice.…”

“…Block-based methods typically entail reduced computational costs, but are also less robust to transformations such as scaling and rotation. Finally, global features, such as the color histograms [17], [18] are the most computationally efficient, but typically have less discriminating power than keypoint-based and block-based descriptions and poor tolerance to global modifications, such as global color changes.…”

“…Here, we use the term dense to refer to a video signature which describes every frame in the video, as in [2], [13]. Conversely, sparse video signatures operate at the keyframe level, with keyframes typically extracted at the shot level [11] or at fixed intervals [18].…”

Perceptual similarity based robust low-complexity video fingerprinting

“…Euclidean distance is a popular distance measure, but fails when the artifact is heavy and localized. More sophisticated distance measures such as Hausdorff distance, partial Hausdorff distance [9] and its variant proposed in [10] outperform Euclidean distance when the query length is short. However, the Hausdorff based distance measures are computationally expensive because they are designed to work well in very difficult, often impractical cases such as when the video frames are permuted.…”