Picture and Video Files

By Timothy Everingham, TUGNET (The User Group Network)

In regards to the size of picture, graphics, and video files, when you try to compress files there reaches a point where you start to lose something.

First you have to understand that if it is either an image or video file each little pixel or time interval is represented by bits of data. You may have heard 8, 16, 24, 32 bit color. The number of bits represents the color of each pixel. On an 800x600 image you have 480,000 pixels (800 multiplied by 600). Then you multiply the color bit depth to get the size of the image representation (800 x 600 x 24 bit equals 11,520,000 bits). Divide by 8 and you get the number of bytes (11,520,000 bits divided by 8 equals 1,440,000 bytes). It get worse with video because you have to multiply by the frames per second and by the number of seconds in the video (at a standard 30 frames per second it would be 41 Megabytes per second or 2.4 Gigabytes per minute). Now if you have a lot of large image or video files you are trying to transfer through a dial up connection to the Internet you have problems. If that is the case, you just figure you will save and send things using a smaller image size and blow up when it will be used. The problem is that when you go to a smaller image size you loose detail in the process. Then when you try to blow it up again it just gets blocky. The same thing goes for video files, the smaller the file size the less detail and nuances are in the file. This really shows up when you take a small image and then try to project it using a good LCD projector. So if you want to project it, show it on a large monitor, or print it out large you need to keep the image size up. This is one reason there is so much hype on the amount of megapixels a digital camera has.

So how then do we get the small file size of images we do. We use compression algorithms, also known as codecs (compressor/decompressor). There are lossless and lossy compression algorithms. But how can there be lossless compression? It is because nature is filled with patterns and uniform things, which is what we take images of. So if we record info about the pattern instead of each pixel of it a lot of file space is saved. An example would be if a person in your image is wearing a blue sweater. Instead of recording every pixel individually you record this area of the image, where the sweater is, in blue. The Tiff image format (.tif) uses only lossless compression techniques, so it is a good format for achieving purposes or something you may wish to edit later. But then there is lossy compression. You lose some of your detail, but a lot less than if you just shrank the image. One of the regularly used lossy compression techniques is color palette reduction. With 24 bit color you have 16 million colors available. However, there are far less that many colors in an image. It takes less space to analyze which colors are in an image, and create a custom color table that will represent all the color in the image in less than 24 bits, write the color representation of the pixels with the custom table code, and then include a copy of the table in the file. This can be used as a lossless technique, but usually this is taken one step farther. In the analysis of the image it is detected how often each color is used and ranked. As the compression rate increases then thenumber of colors that are less used are converted to colors that are close to it but are widely used in the image. This compresses the size of the color table, which decreases the bits per pixel representation of a color used and the blending to nearby colors increases the space that areas of the image that can be described as having one color (this is a situation of compression techniques complementing each other). Some formats start out using only lossless compression techniques but as the level of compression increases they start to use lossy compression. Some common formats that use a combination of lossless and lossy compression are GIF (.gif) and JPEG (.jpg). Remember once you lose image data using a lossy technique you will not have it anymore unless you still have the original or a lossless copy. The compression methods discussed so far are referred to as spatial compression because they reduce the file size by compacting the description of the visual area (space) of an image.

Video is made up of sequential images played over time. This means we can not only use spatial compression; but also use temporal compression, which compresses the file using analysis of the difference and similarities of the frames in an image. This usually exhibits itself by the first frame in a video sequence being compressed just like a still image using spatial compression, but in the following frames only the changes from the previous frame is recorded. Periodically or when there are major changes between frames, a new initial frame, also called a keyframe, is produced followed by more just recording of differences between frames. These temporal compressions can be lossy or lossless. As with still images you should achieve them using only lossless compression. However, using only lossless compression may be impractical because of the limitation of your computers or hard drives’ speed or hard drive space, but keep lossy compression at a minimum.

If you are creating a video DVD for achieving that uses the MPEG-2 format that uses both lossy and lossless compression techniques, it is better to achieve the file using the DV codec it was originally imported in from your video camera and writing that file to a data DVD or CD. Be aware that some compression techniques and settings are better for some types of video than others. If you just have a talking head in your video with a static background you can compresses it greatly without degrading the image much, But if there is a lot of fast action or panning of the camera, the amount of compression you can successfully have you have to be more picky on what codec you should use.

File size matters because it limits what you can use the file for. If you just want a small file you can send to a friend’s dial up connection or have friends using a dial up connection view via a website a small file is good. However in printing large photos, showing on a large computer display or having it projected on a screen small files are not good. Use only lossless compression techniques on things you want to achieve or edit later because it keeps your options open. Keep your archival file large and make smaller copies of it for other purposes.

It is fortunate that hard drive and DVD and CD blank prices have come down a lot so doing this does not cost a fortune, but not planning ahead on image size and which compression techniques can cost you not being able to fully use your precious digital photos and videos.

Timothy Everingham is Vice Chair of the Los Angeles Chapter of ACM SIGGRAPH, the largest chapter of the Association for Computing Machinery’s (ACM) Special Interest Group on Computer Graphics and Interactive Techniques and one of Southern California's significant professional organizations within the entertainment and media industries.

http://home.earthlink.net/~teveringham