Video that is digitized from film has a peculiar noise known as a "film grain noise." This type of noise makes digital video considerably less compressible and may hurt subjective video quality (some viewers prefer film grain noise; others do not). On the other hand, artificial or reconstructed film grain noise can make compression artifacts less noticeable and make the video livelier, and it can improve the visual quality. Recently, considerable attention has been paid to film grain noise removal (degrain) and reconstruction (regrain); degrain and regrain features are currently included in the H.264 video coding specification.

We offer a complete degrain/regrain technology that can be used for detection and suppression of film grain noise, as well as its reconstruction and synthesis. Denoising and reconstruction can be performed using several approaches, each with its own advantages.

Degrain/regrain technology can be used for simulation of professional film tape when using a lower-quality digital source. Another use of this technology is simulation of a uniform film type for display of video that may have been captured from a number of different sources, such as rendered video.

Since suppression of film grain (with grain parameters estimation) and application of film grain are two separate processes, it is possible to use this technology for the removal of any type of noise or to use it as a separate tool for the application of film grain to any input video.

The algorithm can be applied to both moving and static pictures.

Primary Applications

• DVD and HD DVD recorders
• Video CD/DVD/DivX players
• Digital camcorders
• Television equipment
• Stream pre-processing for better quality/compression ratio trade-off after encoding by either lossy or lossless video codecs
• Professional video remastering
• Home video processing

Key Features

• One-pass real-time processing
• Several methods, with various trade-offs, for grain noise removal
• Several approaches to grain reconstruction (generation) including analytical models and a film grain samples database method
• Application of grain noise patterns extracted from any video
• Applicable both to video and to static images
• Suitable for hardware implementation

Basic Deliverables

• Source code for a reference implementation in C
• C and assembly language source code for an implementation optimized for the PC (if required)
• Algorithm description
• Software description
• Verification instructions