Twilight of the Pixels - Shifting the Focus to Vector Graphics
Although an experimental vector video codec might foreshadow a revolution in video scalability and definition, the more immediate outcome will likely be a dramatic increase in encoding efficiency.
A pixel, by nature, is a part of a larger picture. The smaller the pixel is, the more of them that can compose the larger, complete image (and thus, the higher the definition). The finer edges give the picture more resolution, as the higher definition allows a more faithful image. We have seen resolution become finer and finer over the years, which is basically the result of a greater capacity for smaller pixels as digital graphics evolve. But what if pixel size and quantity were no longer the deciding variables in the quality of an image? What if images could be rescaled with little to no loss in resolution?
What Are Vector Graphics?
Vector graphics used to be the personal computer’s primary display system. In contrast, pixel bitmaps (also known as rasterized images) were developed in the 1960s and '70s, but did not come to prominence until the '80s. Since then, pixels have played a huge role in how we create and consume photography, video and a great deal of animation and games. Nevertheless, vector graphics have been employed in digital visual design over the years, and their influence broadens as the technology improves.
As opposed to rasterized images (which map out individual color-valued pixels to form bitmaps), vector graphics employ algebraic systems to represent primitive shapes that can be infinitely and faithfully rescaled. They have evolved to serve various computer-aided design applications, both aesthetic and practical in purpose. Much of vector graphics technology’s success can be attributed to its practicality — as rescalable graphics have many uses across various technical vocations. Generally speaking, however, their ability to depict photorealistic, complex visual presentations is lacking in comparison with the rasterized image.
Traditionally, vector graphics have worked aesthetically where simplicity is virtue — such as in Web art, logo design, typography and technical drafting. But there also exists recent research into the possibility of a vector video codec, which a team at the University of Bath have already begun to develop. And although the implication may be a form of video with augmented scalability, there are other possible benefits, as well as limitations, to explore.
Vector Video Codec
A codec, by nature, encodes and decodes data. The word itself variably serves as a portmanteau of coder/decoder and compressor/decompressor, but both refer to basically the same concept — the sampling of an external source reproduced in a quantized format. Video codecs encase data that determine audio-visual parameters such as color sampling, spatial compression and temporal motion compensation.
Video compression largely involves encoding frames with as little redundant data as possible. Spatial compression analyzes for redundancy within single frames, while temporal compression seeks to eliminate the redundant data that occurs among image sequences.
A large part of vector graphics’ advantage in video encoding would be its economy of data. Rather than literally mapping out images in pixels, vector graphics instead identify points of intersection along with their mathematical and geometric relationships with one another. The “paths” that are thereby created generally provide for smaller file sizes and transmission rates than a pixel map would if the same image were rasterized, and they don’t suffer from pixelation when scaled up.
The first thing that seems to come to mind when considering a vector video codec is the (perhaps a bit quixotic) concept of infinite scalability. While I believe that a vector video codec could facilitate scalability that is dramatically augmented in comparison with rasterized video, image sensors (such as CMOS and CCD — the two dominant image-sensing devices found in modern digital cameras) are pixel-based, so rescaled picture quality/fidelity would taper off at a certain threshold.
A vectorized rendition of an external source image is achieved by way of a process known as autotracing. While simple shapes and paths autotrace easily, complex color shades and nuances have never translated easily as vector graphics. This creates an issue with encoding color in vector video, however color tracing in vector graphics has made significant strides in recent years.
Beyond the image sensor and the video codec, the next important link in the chain is display. Early vector monitors used cathode ray tube technology similar to those used for rasterized picture, but with different control circuitry. Rasterization is the dominant modern display technology. In the visual effects industry, there is a process called "continuous rasterization" that interprets vector graphics' rescaling in a perceivably lossless manner — effectively translating encoded vector formats’ rescaling capability to a rasterized display.
But no matter what the codec or display; the best, most detailed picture can only come from a quality source. Vector video encoding could drastically improve video scalability, but only to the extent of the source’s quality. And the source is always a quantized sample. But if the vector video codec does not swiftly incite a revolution in video resolution and scalability, it may at least offer high-quality video with significantly less cumbersome encoding.