NVeiler Video Filter

Plug-in for VirtualDub that hides faces and objects

Automatically detects faces in a frame and hides them by using pixelation effect. Also it is capable to track the moving faces in subsequent frames and hide them, as well as detect new faces without user intervention.

Users can choose to hide selected faces or parts of the faces, as well as use the plug-in for hiding other objects.

Technical Details

  • Configurable automatic face detection – Users can adjust three parameters to improve face detection quality: face confidence threshold, smallest distance between eyes and largest face roll angle. See the plug-in's documentation for more information on these parameters and their values.
  • Supported video resolution – There are no general limitations on video size or aspect ratio. However we recommend to provide video files with at least 640 x 480 pixels (or at least 0.3 MegaPixels for other aspect ratios) resolution for confident face detection and subsequent tracking.
  • Frame processing performance – An HD video (1280 x 720 pixels) file with several faces in a frame and 15 degrees face roll tolerance is processed at a speed of about 12 frames per second on a PC with Intel Core 2 processor running at 2.66 GHz. The performance is mostly dependent on a frame size and the chosen face detection parameters. The number of faces in a frame has no significant influence on the performance.
  • Technology background – The plug-in is based on VeriLook biometric face recognition technology and SentiSight object recognition and tracking technology. The VeriLook-based biometric products have millions of deployments worldwide including access control, identity verification and other security-related applications. The SentiSight is intended for computer vision applications that are able to find the required objects in videos or still images.