Sunday, September 25, 2011

Does converting to ProRes really matter?

It started with a simple conversation between myself and another filmmaker after a day of shooting on the Canon 7D. We discussed the advantages of transcoding H.264 footage for faster and easier editing as well as improved quality in color correction. But a thought came to mind following this little chat: how much does this really show in the final product? Ever since DSLR cameras started to become popular for film making, transcoding to ProRes has always been a given, mostly because of advice passed down from respected authorities in the community.

The footage we shot that day was captured using the Technicolor Cinestyle profile, and would require considerable color correction in the blacks to suppress some fogging from direct light hitting the lens (a side effect of capturing some great flares.) It was during initial color correction tests on the footage that I decided to compare transcode ProRes versions of the footage to the original H.264 clips. Even though I have worked with this post workflow hundreds of times, it wasn't until doing this simple test that I could really appreciate what is gained in the translation between codecs.

Here is the basic breakdown of the workflow in this comparison. Footage was shot on a Canon 7D at 640ISo in native H.264. Using MPEG Streamclip, the clips were transcoded to ProRes422 (HQ) at 100% quality. In Final Cut Pro the converted clip and original clip were placed into a ProRes timeline and the S-Curve Technicolor LUT was applied using LUT buddy. A 3 way color correction was then applied to achieve the specific look in the shadow (in the case of this shoot, a silhouette). Lastly, still images were exported as uncompressed tiff. For comparison sake, the images were composited together in Photoshop and saved as a level 12 JPEG. (While I have taken great efforts to avoid introducing additional compression to the images below, Blogspot does recompress the uploaded images, so some aspects of my depiction may be harder to distinguish online compared to the source image.)

The image above shows a cropped portion of the shot uncorrected. Click on the image to view at full scale. At first glance there are few major differences between the shots. Upon closer examination we see smoother gradations in the color flare at the top of frame, and reduced noise in grays and blacks.

The corrected version of this shot holds true to the uncorrected version, except we can now see the strength of ProRes to hold up to severe color shifts. A noteworthy part of this image is the light source in the lower left frame. The flare rays in the H.264 image show considerable stair stepping and artifacting, while the ProRes image maintains smoother gradations. These differences are critical when performing extreme color correction or when compositing chroma key images, which tend to suffer stair stepping on edges.

These differences in color gradation can also be seen using Final Cut's built-in scopes. This can be seen best in the histogram (top right) and the RGB parade (bottom right.) The H.264 image appears in the scopes as a banding of colors, while the ProRes image shows smoother curves.

These findings may not be revolutionary, but I do feel they are reassuring. This reenforces that the workflows we filmmakers have established are worth the extra time and storage space, because the final product is superior both visually and technically. Also keep in mind that the advantages of transcoding ProRes are not limited to only DSLR footage but can be applied to other formats that suffer from MPEG compression such as XDCAM or AVC.

No comments:

Post a Comment