So instead of x-y coordinates, let's say we have frequency coordinates. Now, some imagination! Imagine you can transform any dataset that varies over space(or time) - something like the brightness value of an image, into a different coordinate space. You can easily transform between the two systems. You can represent data in some fundamental units. Now that you understand information entropy, let's move on to transformations of data. This type of encoder is called an entropy encoder - it's a general-purpose lossless encoder that works for any type of data. The information entropy in this dataset has not changed - you've just converted between representations. This is obviously an oversimplification, but you've transformed some data into another shorter representation of the same information. You would just say '10 tosses, all heads' - bam! You've just compressed some data! Easy. How would you describe this information to someone? You wouldn't say HHHHHHHHH. Suppose you have some strange coin - you've tossed it 10 times, and every time it lands on heads. If you have record two coin tosses, you'll need 2 bits. It is minimum number of bits that must be used to represent all the information contained in a dataset.įor example, if your dataset is the result of a single coin toss, you need 1 bit of entropy. Note that it is not simply the size of some dataset. Information entropy is the number of bits required to represent some information. Now we're getting to the juicy bits! Ha puns! If you took an information theory class, you might remember information entropy. Imagine your car weighed just 200 lbs!ħ% wow! How do you discard detail information like that?įor this we need a quick math lesson. The image on the right weighs in at 7% the size of the original - and we haven't even compressed the image in the traditional sense. See how the compressed one does not show the holes in the speaker grills in the MacBook Pro? If you don't zoom in, you would even notice the difference. Here is a close-up of the original compared with the image post-discard. H.264, like other lossy image algorithms, discards detail information. Or maybe we can crop out a thick border around the edges of the frame, the important stuff is in the middle anyway. We would use only 3/4th of the space we need. So maybe we can zero out those pixels and discard that quadrant. Maybe the top right quadrant is useless all the time. There are few obvious ways to trim out images. Important bits? How does the algorithm know what bits in my frame are important? Bit for bit, the original source image can be recovered from a PNG encoded image. H.264 is a lossy codec - it throws away less important bits and only keeps the important bits. This concept of throwing away bits you don't need to save space is called lossy compression. You remove everything except the things that matter. What is the first thing you do? You shed some weight. Imagine you're building a car for street racing. There are very many tricks! H.264 uses all the tricks you can think of (and tons you can't think of). How is this even possible? All right, what's the trick? So H.264 would seem to be 1500x as efficient as PNG. It looks like we're storing 300 times the amount of data in the video. A single frame of that video in PNG is 1015KB. The H.264 video, 300 frames long is 175KB. 5 Second 60fps H.264 video of the same Apple homepage 175KBĮh.

PNG screenshot of the Apple homepage 1015KB.I captured the screen of this home page and produced two files: Even SSDs have trouble dumping this straight from RAM to Disk. A 50GB Blu-ray disk will only hold ~2 mins.

Each pixel takes 3 bytes to store - one byte each for the three primary colors (red, green and blue).ġ080p 60 Hz = 1920x1080圆0x3 => ~ 370 MB/sec of raw data. So it's a 3D (2 spatial dimensions and 1 temporal) array of bytes. Also note that many of the concepts explained here apply to video compression in general, and not just H.264.Ī simple uncompressed video file will contain an array of 2D buffers containing pixel data for each frame. This post will give insight into some of the details at a high level - I hope to not bore you too much with the intricacies. It is the result of 30+ years of work with one single goal: To reduce the bandwidth required for transmission of full-motion video. H.264 is a remarkable piece of technology. It is ubiquitous - internet video, Blu-ray, phones, security cameras, drones, everything. H.264 is a video compression codec standard.

H.264 video compression