CineForm Compression Subsystem
Video file formats use standard wrappers like AVI, QuickTime (MOV), or MXF to ensure broad playback compatibility with numerous video applications. Within the AVI or MOV wrapper, the underlying file might be compressed (or uncompressed) using various codecs, all of which guarantee compatibility with a calling application through OS-recognized AVI or QuickTime interfaces. The primary function of a codec is to ensure that the decoded image best represents the original uncompressed source – within the inherent constraints of the chosen codec and the parameters selected during the encode (or compression) operation.
Codec designs are many, as are the resulting visual fidelity and CPU requirements that result from various designs. As a general rule, most codecs have been designed for either acquiring images or video inside a camera (JPEG, DV, DVCPRO, AVCHD, DNxHD) or for distribution (MPEG, h.264).
CineForm has taken a different approach with its compression engine – instead designing a compression subsystem specifically optimized for the needs of post production, and with the inclusion of many system features. To be clear, there is a codec at the heart of CineForm’s compression subsystem, but the codec is surrounded by additional processing engines that operate underneath the AVI/MOV interface layers through which calling applications access compressed data. This allows CineForm’s compression subsystem to perform myriad system-level features difficult to otherwise implement.
Following are the major component pieces in a CineForm compression subsystem:
CineForm Wavelet Codec: The design goals for CineForm’s codec were:
1) Maintaining extremely high visual fidelity through a rigorous multi-generation post-production workflow, including high bit precision yielding high signal-to-noise ratio.
2) Offering extremely fast performance on Intel architecture CPUs such that multiple streams can be processed in real time without the need for specialized hardware.
The design characteristics of CineForm’s codec are the following:
Transform: Full-frame Wavelet. “Full-frame” means the entire image is transformed without breaking the image into “blocks” which are the basis for DCT-style transforms (JPEG, MPEG, AVCHD, etc), and which provide the ubiquitous artifacts often visible in complex image sequences. CineForm’s full-frame Wavelet transform cannot suffer from these DCT image degradations.
Chroma Formats and I/O Bit Depth: 10-bit YUV, 12-bit RGB(A), and 12-bit CineForm RAW
Spatial Resolution: Unrestricted
Bit Rate: Variable (constant quality)
Encoder Quality Settings: 6 (user selectable)
Symmetry: Symmetric (similar speed for encoding and decoding)
Multi-Resolution Decoding: Wavelet algorithms offer a valuable feature for use in playback or other post-production applications – the ability to decode a full-resolution image to a lower resolution. Specifically the ability to decode to ½ or ¼ resolution in each dimension is allowed, while using less CPU load compared to a full-resolution decode. CineForm software takes advantage of this characteristic dynamically during editorial.
Performance: Function of CPU speed and memory bandwidth, but below are representative examples run on a 3.2GHz Intel Corei7 (single chip quad core):
|10 bit 1920×1080 4:2:2||232 frames per second|
|12 bit 1920×1080 4:4:4 (CineForm 444)||104 fps|
|12 bit 4096×2048 4:4:4 (playback at 2K)||52 fps|
|12-bit 4096×2048 4:4:4 (playback at 4K)||20 fps|
Relative Performance versus JPEG2000: CineForm compression ranges from around 5X to 10X faster than JPEG2000. In other words, compression performed by CineForm in an hour may take up to 10 hours for JPEG2000 to perform on the same imagery.
Visual Quality (versus HDCAM SR): CineForm visual fidelity routinely exceeds that of the respected HDCAM SR format which is a mainstay in professional production and post-production. Read additional quality analysis versus HDCAM SR on StEM test footage….
Visual Quality (versus HDV and DVCPRO HD): CineForm visual fidelity substantially exceeds that of most camera compression formats such as HDV, DVCPRO HD, AVCHD, etc.
Temporal/Spatial Transform Engine: The Transform Engine allows source and output formats to be different than the compressed format, in either spatial resolution or chroma format.
Spatial Transforms: CineForm spatial processing is unrestricted, and handles all standard resolutions: SD, HD, 2K, and 4K and beyond.
Color Space Transforms: Color space transforms can move images between RAW, 4:2:2, and 4:4:4 chroma in real time. Transforms are always performed using floating-point arithmetic to preserve color accuracy. When decoding from CineForm RAW sources, CineForm’s demosaic algorithms are performed in real time.
Active Metadata Processing Engine: CineForm’s Active Metadata Processor allows numerous non-destructive operations to be performed on image frames after decompression including color processing (white balance, gain/offset, gamma, saturation, RGB matrix controls, 3D LUTs and more), image flip, 3D processing, and more. Please read separate Tech Notes that discuss Active Metadata, and 3D processing in more detail.