With all of the web video streaming services at our disposal that can instantly serve up 1080p video, it is easy to forget how much of a challenge it is to store and encode moving images and sound. When we think about analog motion picture, film comes to mind and it is easy to visualize the difficulties of loading, storing, protecting, and projecting it. Even relatively short films are incredibly bulky and difficult to manage, giving us some rationale for why we needed to digitize video.
However, there is a complication: audio and video information, in their raw form, take up huge amounts of memory. High quality uncompressed digital audio is made up of around 44 thousand samples, each usually 16 or 24 bits per sample, for every second of playback. For even a three minute pop song this adds up quickly. Uncompressed video is a collection of pixel arrays, usually 24 or 32 bits per pixel, and when you multiply the resolutions of the image by the 30-60 frames per second that you can expect from video, you will be faced with a massive amount of storage. Even though storage is very cheap these day and companies like Google deal in petabytes worth of storage at a time, files of that size are impossible to send over a network with any kind of real-time reliability. This is where encoding comes in: it has enabled the rich playback abilities that we enjoy today.
Encoding schemes such as MPEG take advantage of the fact we maintain persistence of vision and are more susceptible to high frequency changes in an image sequence. Ever wonder why you will usually see small blocky regions in an encoded video? This is because the image is given a spatial frequency analysis in blocks that are usually 8×8. Converting something such as an array of pixels or audio samples into a sum of cosines (known mathematically as DCT) leads to a representation that is significantly more compact, and in the case of most video there is little change from pixel to pixel. The most current encoding schemes, on top of this type of analysis, can perform recognition of objects in frames and motion estimation so that only changes in the image that are perceptible to the viewer are part of the output. What this means is orders of magnitude less information that must be transmitted in order to play the video at reasonable quality.
This is just scratching the surface of the abilities of modern video encoding, but we have established that it is an essential part of modern media delivery capability and a core driver to value in many businesses. Contact us to learn more about our efficient, high quality enterprise video encoding services.
You may be interested
The Two Dominant Delivery Formats For Low Latency: HLS and DASHTushar Sonal - October 15, 2019
The Two Dominant Delivery Formats For Low Latency: HLS and DASH Video consumption has massively increased in the past decade, thanks to younger demographics spending considerable time…
How to Measure Your CDN’s Cache Hit Ratio and Increase Cache HitsTushar Sonal - September 28, 2019
How to Measure Your CDN’s Cache Hit Ratio and Increase Cache Hits What happens when a user on your web app initiates a web request? The request…
How AI And Machine Learning Are Changing Content Delivery NetworksTushar Sonal - September 22, 2019
How AI And Machine Learning Are Changing Content Delivery Networks Technological innovation has entered into an exciting era — the twin pair of Machine Learning (ML) and…