Dressing for Streaming Success; Brown is the New Black

I'm preparing for my full day seminar entitled Producing a Webcast from Soup to Nuts at Streaming Media East in Manhattan on May 12, 2014. Part of the process involves watching a bunch of live events to glean tips and tricks and mistakes to avoid. Two videos of the same subject that I noticed yesterday illustrated two critical principles of set design and dressing for streaming success, so I thought I would share. 

To explain, one of the most critical components to high quality video is the background used in the video and the clothing worn by the subject. When planning your clothing and background, you should keep the concept of contrast ratio front of mind.

Briefly, concept ratio describes the ability of a system to distinguish detail in the brightest and darkest regions in a frame. Expensive camcorders with large imaging devices have a great contrast ratio, which is why the TV and movies can pull off shadowy scenes that our $3,000-$7,000 camcorders can't manage. Consumer camcorders and webcams have even lower contrast ratios, which typically leads to loss of detail in either the brightest or darkest regions in the frame. You can see this on the left in figure 1, where the tony black shirt looks like a dark blob, and the black fabric seems to suck in all the light in the frame. 

eft-dress4success.jpg

Figure 1. When choosing clothing, mind the contrast ratio.

When planning your background and clothing, you should minimize the difference between the brightest and darkest regions in the frame. On the left, there's a huge difference, and the camera failed to pick up the detail in the shirt. On the right, there's much less--in fact, the subject is wearing perhaps the ideal clothing, moderately colored solids that contrast well with the background. Not to go all geeky on you, but you can see the difference in the brightness extremes in Figure 2, two luminance histograms of the two frames, which shows the distribution of the brightness of the pixels in the two images. 

On the image on the left, you see a huge clump on the left; that's the black shirt; this forces the camera to try and preserve detail in the brightest and darkest regions, which it cannot do; hence the black blob. On the right, you see the bulk of the brightness in the middle, with very few pixels at either extreme. This allows the camera to preserve detail in all relevant regions, producing a much higher quality frame.

To avoid issues like this, when planning your shoot, avoid bright whites and particularly blacks, since they may cause the loss of detail shown in the left in Figure 1. Black may look great at cocktail parties (since your eyes have a marvelous contrast ratio), but it's to be avoided when on camera. 

eft-dress4success-histogram.png

Figure 2. This luminance histogram shows how the image on the left strains the contrast ratio of the camera. 

What else can we learn from this video? Take a look at Figure 3. While bookshelves are great for establishing credibility, books contain a lot of detail that the codec must strive to preserve during encoding. This takes bits away from more important regions, like the face, which could degrade quality there. In addition, the eye naturally notices the loss of detail like that shown in Figure 3, which reduces the overall perception of quality. 

eft-dress4success-bookshelf.png

Figure 3. Avoid unnecessary detail in the background.  

When planning your shoot, choose a background with minimum detail; while bookshelves may look great in the camera monitor, or even in your high-bit-rate streams, someone out there will likely be watching your video on a low bitrate connection on a cell phone, and the video will probably look like a muddled mess. When it comes to the background, if it moves, has excessive details or is too light or dark, it could definitely degrade the overall quality of your video. Because sometimes the best way to produce top-quality streaming video has nothing to do with compression at all.


Comments (4)

Doug
Said this on 4-24-2014 At 06:39 pm

If the camera is fixed relative to the background, and there's no zoom in or out, and the encoding is long-gop, you won't lose too much bitrate to background details, except around the speakers head as it moves left/right/up/down. Around that area, uncovering background detail will suck up bits.

The encoder will also suck up bits to create a fresh reference frame at the start of each gop, so better to keep the gop long. 

It's probably ok to have some detailed pattern near the edges of the frame, and it provides some interest and relief for the eye, but it's better to have a plain background immediately behind the speaker.

Said this on 4-24-2014 At 08:31 pm
Doug:

Great commentary; thanks for weighing in. Have you had any issue with moire effect? I've had horrible results with high-detail wallpaper, venetian blinds, rattan curtains and similar backgrounds. Find detail just scares me in any form.

Jan
Said this on 4-25-2014 At 02:30 pm

What are the encoding implications of a shot where the talent is in sharp focus in front a a green screen keyed over a background is intentionally blurred or out of focus ? With fewer sharp edges would that part of the frame take fewer encoder resources, improving the foreground (talent) detail ?

Jan Ozer
Said this on 4-27-2014 At 07:13 am
Andy:

Sounds like it would, but the funky thing about encoding for streaming is that you never really absolutely know until you try it. So bring a notebook to the shoot; capture some video and try the encode.

Thanks for contributing to the Streaming Learning Center.

Jan Ozer
Post a Comment
* Your Name:
* Your Email:
(not publicly displayed)
Reply Notification:
Approval Notification:
Website:
* Security Image:
Security Image Generate new
Copy the numbers and letters from the security image:
* Message: