For example, i started downloading Community on Netflix, most of the episodes are 21 minutes long, but they range in size from 76 to 116mb. They’ve got nearly identical run times, and therefore similar amounts of frames and audio, so what causes the noticeable discrepancy?
In: Technology
Video is just a large set of images known as frames. Let’s say you set up a tripod and filmed a video of some clouds floating across the sky. If you had to describe every frame, you would be repeating a lot of blue pixels making up the sky! This would take a lot of space to list the blue color for every frame which is why raw video is extremely large.
Video compression methods take advantage of the fact that not many things change frame-to-frame. It takes less information to describe the changes per frame than listing all the contents of those images. All you need is yo describe the edges of the floating clouds. Sometimes, a scene won’t have many changes so you don’t need to save as much data, making the video smaller. On the other hand, a large amount of changes like action scenes or panning shots will require more information to describe, making the video larger.
Video compression often sets a limit on the bit-rate or the amount of information stored per time for the output video. The video is free to go below the bit-rate for a smaller size. People could also set a max video size and make the bit-rate change to fit that size. Then you can fit video files to limited storage media like CDs, DVDs, and BluRays.
The videos are compressed. When you compress something, you look for redundant information and try to reduce it. Let’s say I made a 20 minute video of a white screen… I could record 1920×1080 pixels of white for 60 frames per second over 20 minutes… or I could just tell you that all the pixels are white for 20 minutes, and you’d get the same result. Or I could have a 20 minute video of a photo of a cat. I could send you the full data for it, or I could just send you the one image and then tell you that it’s just that image for 20 minutes.
Obviously most videos aren’t just a white screen or a cat… but think about any time you’re watching a video and the camera isn’t moving… 3/4 of the screen is completely stationary, so they just need to send the data representing what is changing.
Do this for every frame of video and you can have much less data.
Part of video compression algorithm looks at the similarities between frames. If someone is monologuing on screen against a static background, the algorithm is smart enough to know that it only needs to store the visible background elements for one frame and repeat it, rather than a new image for every single frame. The bigger episodes probably have more movement in them.
Latest Answers