The SI-2K with embedded SiliconDVR records directly to the CineForm RAW™ digital intermediate codec, a powerful, 10-bit visually lossless codec that natively encodes the RAW bayer data from the single-sensor in the Silicon Imaging cameras, and preserves the RAW data at the codec level until the data is "flattened" for the final output. With it's 4:4:4 RGB decodes, non-destructive metadata-managed color pipeline, and 32-bit float internal processing engine, the complete tonal range of the image is preserved through multiple passes while its light 5:1 wavelet-based compression avoids any visible compression artifacts.

CineForm RAW™ is a wavelet-based codec unlike the typical DCT-style compressor used in every other tape-based digital camera system on the market. DCT, or Discrete-Cosine-Tranformation, is the basis for such codecs as DV, DVCProHD, DVCAM, MJPEG, MPEG, HDCAM, HDCAM-SR, D5 and JPEG. They primarily work by splitting the image into 8x8 blocks and then quanitizing those blocks with various algorithms to compress the image. During high compression or insufficient data rate to encode a complex image, these 8x8 areas show up as the mosquito noise or "blockiness" that is typically attributed to in-camera compression.

Wavelet-based technology, on the other hand, is one of the most effective methods of image compression available today, and is the basis for both the CineForm RAW™ codec, as well as other next generation codecs such as JPEG2000 which has been ratified by the Digital Cinema Initiatives commission (DCI) as the standard codec for digital cinema projection. Wavelet algorithms are based on the concept of multi-resolution analysis, where the entire image is analyzed and then compressed, rather that being split into small chunks where compression boundaries can become visible. As a result, wavelet compression takes on a more "natural" appearance, with high compression ratios gracefully "softening" high-frequency details rather than producing un-natural and highly visible block-style patterns. For a given compression ratio, wavelets can maintain a much more pleasing and natural appearance to the subject image than other DCT-style codecs are capable of.

Roll-over image to see comparison between DCT and Wavelet

The high compression ratios in this example exaggerate the differences in the DCT and wavelet compression methods, demonstrating how wavelets maintain a more "natural" appearance without exhibiting traditional DCT macro-block patterns, gradient banding, or mosquito noise, even at high compression ratios.

For more information about CineForm RAW™, please download a detailed white-paper here.

 

®Microsoft Windows XP, Adobe, Adobe Premier Pro, Cineform, ProspectHD, Silicon Imaging. All Rights Reserved. For more information please email us at HD@siliconimaging.com