D800 Bitrate Evaluation

[LPowell]

Here are results from comparison tests of the D800 in 1080p30 video mode, using the stock 24Mbps bitrate, and hacked 36Mbps, 54Mbps, 64Mbps bitrates. I took these shots with a Tamron 24-70mm f2.8 lens set at 70mm, f5, and ISO 100. The camera was locked down on a tripod and used the Neutral Picture Profile. I selected a sunlit scene of a running stream of water with highlights on the water ripples. This type of subject presents a myriad of high-contrast moving images that tax the motion tracking and bitrate allocation algorithms of the camera's H.264 video encoder.

In 1080p30 mode, the D800 records in an IPB format using a 15-frame GOP. After loading the video files into a 32-bit Adobe After Effects project, I selected I, P, and B-frames from the second GOP in each video sample. I then displayed and cropped them at 200% magnification in side-by-side 24Mbps versus 36Mpbs and 54Mbps versus 64Mbps comparison shots, and saved screenshots of each comparison.

I-frame Comparison @ 1080p30 - 24Mbps vs 36Mbps


I-frame Comparison @ 1080p30 - 54Mbps vs 64Mbps


P-frame Comparison @ 1080p30 - 24Mbps vs 36Mbps


P-frame Comparison @ 1080p30 - 54Mbps vs 64Mbps


B-frame Comparison @ 1080p30 - 24Mbps vs 36Mbps


B-frame Comparison @ 1080p30 - 54Mbps vs 64Mbps


H.264 encoders divide each video frame into horizontal and vertical rows of tiles called macroblocks. Sophisticated compression techniques are used to compress the encoded data within each macroblock, reusing image data from adjacent macroblocks and video frames. After the H.264 decoder reconstructs the encoded pixel data, it uses a built-in Deblocking Filter to blend the rectangular edges of the adjacent macroblocks together. With extremely detailed moving images, however, the encoder's image quality may deteriorate when it reaches the limits of its maximum bitrate. This can produce not only smeared image details, but undesireable macroblock artifacts as well, which can be seen in the sporadic rectangular edges in detailed areas of the decoded frame grabs.

Frame grabs from the 54Mbps and 64Mbps videos show significantly fewer macroblock artifacts and finer image details than the 24Mbps and 36Mbps videos. However, macroblock edge artifacts can still be seen in 54Mbps and 64Mbps frames. More seamless results could potentially be produced by hacking the encoder's Deblocking Filter coefficients. A more effective Deblocking Filter might also make it possible to achieve artifact-free image quality at bitrates low enough to be played back in-camera.

[LPowell]

Here are download links to the original, unedited sample video files examined in the first post in this thread:

D800-30p-24Mbps.MOV: https://vimeo.com/90937884
D800-30p-36Mbps.MOV: https://vimeo.com/90937885
D800-30p-54Mbps.MOV: https://vimeo.com/90937883
D800-30p-64Mbps.MOV: https://vimeo.com/90937886

[Samuel H]

great test

[max]

Moved to Showcase, thanks for the detailed test results!

[JLM]

Thanks for explaining how higher bit rates improve the image quality.If I am recording to the Ninja2 from my D800, will I still see an improvement to the image.

[geoffcbassett]