The most obvious symptom of the lack of spatial resolution in a display is the appearance of jagged edges or 'jaggies'; that is those disconcerting little jumps that occur in lines that are drawn close to the vertical or horizontal. As might be expected, jaggies can be explained in terms of a sampling process. A straight line or edge is not a band-limited phenomenon; but contains frequency components that extend to infinity which as we already know, is forbidden in a sampled data system if aliasing is to be avoided.

Figure 8 Removing jaggies

As jaggies are an aliasing phenomenon, the obvious way to improve or remove them is to increase the sampling rate i.e. the spatial resolution of the display. In many systems however, this is not possible as it will already be running at its limit anyway so another scheme is called for. If the sample rate can’t be increased then the high frequency content of the waveform must be reduced by filtering it and the simplest way to do this is to utilise the ability of the stimulus generator to reproduce greyscales. For an  easy way to do this, look at the illustration in Figure 8 which shows a section of a dark bar drawn across a grid of pixels (the larger squares) to which has been added a grid of smaller imaginary pixels in the ratio 1:64. Instead of allocating one of two luminances to the real pixels in the traditional way just count the number of smaller pixels covered and use this number to allocate one of 64 different luminances instead. Although this tends to blur the edge, because we are in effect low-pass filtering it, it also allows the edge to be positioned spatially with far greater resolution than before; a situation analogous to hyper-acuity in the human visual system.

It is an interesting fact though, that many observers state a subjective preference for images with a significant, if spurious, high frequency content over those without, even if the high frequencies come from the raster of a CRT display or random lines drawn on the image.

^ Back to top