The three types of video resolution
When we speak of resolution, we are most likely referring to “spatial” resolution. But a digital video signal actually varies in three different types of resolution: spacial, temporal and color. Of these, color resolution, or “bit depth” is easily the most misunderstood. Let’s quickly review them all, and pay special attention to the last.
Quite simply, the number of pixels in a given frame. Sometimes expressed as width x height (1920×1080), sometimes expressed as “megapixels” (12 MP) and sometimes expressed as shorthand for a standard (4K).
More spacial resolution can reveal more detail, up to the limits of the human visual perception. (Only so much detail can be perceived based on the size of output and viewers distance from output.)
Humans are more sensitive to horizontal resolution that vertical resolution.
Acquiring video at more spacial resolution than is strictly necessary for output allows for flexibility in post (cropping and zooming) and a general increase in perceived quality (supersampling, lack of resolution artifacts.)
The number of frames per second. Usually expressed as frames per second (24 FPS) but sometimes also expressed as a “refresh rate” (60 hz).
Higher temporal resolution playback yields smoother, technically more accurate, motion. Interestingly, though, while more spacial resolution is almost always considered a good thing, 100 years of cinema has caused us to associate lower temporal resolution with higher quality (film versus video.) There are some valid reasons for this, but they are outside the scope of this post.
Strictly speaking, humans can’t see motion. Our brains put together a series of still images, and we are not particularly sensitive to temporal resolution. In the hands of a skilled animator, even 12 frames per second can look very smooth. (Early Disney movies.) Most people cannot perceive an increase in temporal resolution above 60 fps.
Acquiring video at more temporal resolution than is strictly necessary for output allows for slow motion playback when those extra frames are displayed at the same rate.
The number of discrete values between black and white. Combine multiple channels (R, G, B or YUV) to yield “full” color. Expressed as “bits” (8-bit). Each increment yields twice as many steps, so a 10-bit image has 4 times more color resolution than an 8-bit image.
The vast majority of display devices, including professional display devices, can only display 8-bits of color resolution. Film, or the digital projectors in theaters are capable of displaying higher color resolutions.
Human sensitivity to color resolution varies based on brightness. We are relatively bad at discerning fine differences in color resolution in shadows and highlights, and much better at doing so in the middle range. This limitation is exploited both when compressing video (shadows are compressed much more than other areas of the image) and to squeeze more dynamic range into an image (a LOG gamma curve.)
Acquiring video with more color resolution than is strictly necessary for output allows for greater flexibility in post (extreme color manipulation) and a limited increase in perceived quality from supersampling.
Practical implications of color resolution
Because color resolution is most often misunderstood, it is surrounded by the most voodoo and misinformation. The practical answer to “Is 10-bit better than 8-bit?” is not “Yes.“, it’s “It depends.”
1024 (10-bit) tones are always technically better than 256 (8-bit). All things being equal, there’s no reason to not record 10-bit if you have the capability. But all things are rarely equal in the real world. Usually there is a trade-off. With the GH4 (the camera I currently use and most people reading this blog are interested in) that trade-off is an external recorder that adds significant cost (at least double) and bulk (more than double) and reduces convenience (internal SD cards.) Because of these trade-offs, one must ask themselves, “Do I need more color resolution?” The answer is often, “No.”
The reality is, for most real world images, the visible difference is negligible. Spacial resolution trumps color resolution, and usually there is enough spacial detail in an image to hide the lack of color detail. There are certainly exceptions, though.
The most common problem scenario for 8-bit color resolution is when shooting the sky. Subtle gradations in a cloudless sky will reveal the limitations of an 8-bit image. If the area is large enough, visible “banding” or “steps” could be revealed. Any large, flat, monochromatic surface could be problematic…especially if you attempt to radically alter the image in post.
“Dithering” is a common technique for hiding a lack of color resolution, or simulating more color resolution. (For instance, a magazine or newspaper only has 1-bit of color resolution.) By adding a small amount of noise to a video signal (such as film grain) color resolution artifacts can often be hidden.
The human eye is not very sensitive to subtle changes in luminance in highlights. Logarithmic encoding schemes make use of this phenomena to squeeze more dynamic range into a signal than otherwise would be possible.
Color Resolution FAQ
Q: Do I need more than 8-bits to do green screen work?
A: No, but it helps. A well lit 8-bit key will always outperform a poorly lit 10-bit key.
Q: Do I need more than 8-bits to do color correction?
A; No, but it helps. You can push a 10-bit image farther before you are likely to see visible artifacts.
Q: Do I need more than 8-bits to use a LOG (logarithmic) profile?
A: No, but it helps. LOG profiles sacrifice color resolution in the shadows and highlights in order to squeeze in more total tonality (dynamic range.) We’re generally not sensitive to fine differences in these areas, so the “compression” usually goes unnoticed, but see the answer above.