Introduction to Video CODECs : MJPEG, MPEG-4, H.264

CODECs are a critical element of choosing, designing and using video surveillance systems. CODECs can lower the price of overall systems and increase the usability of systems. As such, having a basic understanding of what a CODEC is and why CODECs are used is important.

Fundamental Principle of CODECs

The most important factor to understand in video CODECs is that CODECs help balance off different costs.

For instance, let's say you want to go to the mall and to the supermarket. A few years ago, when gas was cheaper, you might have done this in 2 separate trips. Now that gas prices have increased dramatically, you might want to combine those trips. What's happening here is that as gas has become more expensive, you are willing to trade off lower convenience for savings in cash.

Likewise, using CODECs is a balance between the cost of storage, bandwidth and CPUs. Specifically:

As such, selecting a CODEC always requires you to understand the tradeoffs in cost between using less bandwidth and storage or using less CPU cycles. Generally CPU cycles are cheaper than bandwidth and storage so more advance CODECs save you money. Sometimes, CODECs can be too demanding, especially with megapixel cameras and can potentially cost you more in CPU than you save in bandwidth and storage.

Please read my basic bandwidth tutorial for a review of bandwidth's impact on video surveillance.

CODECs Overview

Video must be digitized for it to be used and viewed on a computer. CODECs are means or choices in how we make the video digital.

CODECs or compression / decompression technologies are used to modify the video that is being digitized. Similar to how you might ZIP files on your PC, the video is compressed on its way into the computer. And just like with opening a ZIP file, the video is decompressed before you use or view the video. Unlike ZIP files, the compression of video losses some of the information (engineers refer to this as lossy compression). However, with the appropriate settings, a user cannot tell the difference visually.

Just like in the movies or TV, video is a series of images that are displayed rapidly one after the other. In the US, TV consists of displaying a series of 30 images per second. When we view these 30 images per second, it's “video” and it looks smooth. The fact that video is made up of a stream of images is quite important for understanding CODECs.

When you use a CODEC, you can compress the video in two fundamental ways:

• Compress the individual image by itself
• Compress a series of images together

When you compress an individual image by itself, you simply take the image, run the compression and output the saved file (technically called intraframe compression). Just like when you use Microsoft Paint and save as a JPEG, video compression of individual images works quite similarly. The difference with video is that you need to do these for a continuous stream of images. As such, rather than simply being a JPEG, it is called Motion JPEG or MJPEG.

The benefit of MJPEG is that it requires very low CPU use. The downside is that storage and bandwidth use can be quite high.

When you only compress an individual image, you ignore what's going on between multiple images in a sequence and often send redundant information. If you are streaming video at multiple frames per second, you often are sending basically the same image over and over again. This can be quite wasteful. It's similar to someone calling you up every minute to tell you nothing changed. It would be far better for the person to only call you when news occurred. You can simply assume during the rest of the time that the status is the same.

When people talk about the benefits of MPEG-4 and H.264, not sending repetitive information is the core source of their strength. Evey so often these CODECs will send a whole image. The rest of the times they only send updates describing what parts of the image have changed (technically called interframe compression). Since it is common that large parts of the image remains the same, this can result in very significant reductions in storage and bandwidth. For example, where MJPEG may send image after image at 100 KB, codecs like MPEG-4 or H.264 may send the first image at 100 KB but the next 3 or 4 images at only 10 KB each. This can approach can reduce bandwidth and storage use by 50 – 90%.

The downside with this approach is that it takes more work for the computer to do this. When you are simply compressing individual images, you do not need to worry about what happened before or what the next image will contain. You simply apply the compression rule and execute. With MPEG-4 or H.264 you need to examine groups of images and make complex calculations of what changed and what did not. You can imagine this can become very complicated and consume lots of CPU resources.

H.264 and MPEG-4 are similar in that they both reduce bandwidth and storage by examining groups of images when they compress video. A key difference with H.264 is that it uses much more complex and sophisticated rules to do the compression. Because H.264's rules are more sophisticated, they can reduce bandwidth and storage even more than MPEG-4. However, the trade-off is that it takes more CPU cycles to do it.

Looking at Current Video Surveillance Systems

The general trend in video surveillance has been a continuous movement to CODECs that save bandwidth and storage. Historically, you have seen products move from MJPEG to MPEG-4 to H.264. The reason why this has happened is because the cost of CPUs to compress the video has decreased faster than the cost of bandwidth and storage. Most experts expect this trend to continue.

Recently, the biggest challenge using CODECs in video surveillance systems has occurred with the rise in megapixel cameras. For years, the maximum resolution of security cameras was constant. All of a sudden with megapixel cameras, the resolution of security cameras has increased by 400% to 5000% or more. The greater the resolution, the harder the CPU needs to work and the more cycles that need to be allocated.

The huge increase in resolution is a little similar to the jump in gas prices. It has changed the economics of CODECs. Whereas historically, for standard definition security cameras, CPU cycles were cheaper than bandwidth and storage. Now, since so much more CPU cycles are needed, it can cost way more in CPU than what you save in bandwidth and storage. As such, most commercial megapixel cameras use MJPEG, especially if they are multi-megapixel (more than 1.3 MP).

One of the most important elements in the next few years will be the development of new approaches and use of new CPUs to reduce the cost of using H.264 for megapixel cameras. Much like alternative energy development hopes to bring the cost of energy down, new approaches are being sought to reduce the use of CPU cycles in compressing megapixel camera feeds.

Conclusion

Understanding the basic choices in CODECs and rationale for choosing CODECs is a key element in video surveillance systems. Please share your questions or feedback below.