|
|
 |
Comparisons |
| |
Video identification has been a topic of active research in academia for more than a decade. While some research findings promise to have the same level of robustness and discriminative power as VideoDNA™, they fall short in practicality due to the complexity in algorithms that make it difficult to implement efficient match and retrieval in a large-scale database. VideoDNA™ presents significantly higher performance compared with those traditional techniques in video identification.
| |
Watermark |
Cryptographic Hash |
VideoDNA™ |
| Does not alter content |
No |
Yes |
Yes |
| Work for legacy content |
No |
Yes |
Yes |
| Robust to content manipulation |
Somewhat |
No |
Yes |
| Identify part of content |
Yes |
No |
Yes |
Watermarking
A watermark is a visible or invisible pattern that can be embedded in video for verification and tracing purposes. While both VideoDNA™ and watermark can be used to identify video content, they are fundamentally different in the way they work and their effectiveness. First, watermarking alters video content while fingerprinting does not. Second, a watermark must be embedded in video before its release in order for the video to be identified by the watermark, while the VideoDNA™ can be computed from video at any time. Most video content that can be found to date, including TV programs, movies and video released in DVD and VHS, and Web video, are not watermarked. Video content from these sources cannot be identified by watermark, but can be identified by their VideoDNA™.
Hashing
A cryptographic hash, also known as a message digest, such as MD5 sum, is often used as a unique identifier for a computer file. A cryptographic hash is highly fragile in that a tiny change (e.g., one bit) in a file will result in huge difference in the hash value. While cryptographic hash is useful in data authentication and integrity check, it is not suitable for identifying video content because the hash value will change dramatically for each slightly different copy of the same video content. On the other hand, VideoDNA™ stays the same for the same video content regardless of the difference in bits order, file size, or file format. In comparison, cryptographic hash identifies the video file, while VideoDNA™ identifies the video content.
Click here for detailed information about VideoDNA™ solutions.
|
|
