Table of Contents | 
Up-conversion: Basic method | Up-conversion: Optimising
Table of Contents | Up-conversion: Basic method | Up-conversion: Optimising
The filter is required to remove as much of the alias signal as possible from the final image. The magnitude response of the filter used in the existing code is shown in figure 1.
Figure 1 - Magnitude response of up-conversion filter
The first filters implemented in Dirac were designed using web-based java applets and MathCAD worksheets. The final version was designed using a 'brute-force' search algorithm written in C++ by Thomas Davies. The filter is a half-band filter that has tap properties that makes it favourable for fast implementations.
The code uses a nineteen-tap filter, which is symmetrical about the tenth tap. Every other tap in the filter equals zero so from the original nineteen taps there are only ten that are used in the calculation. Using the non-zero taps on the original image is equivalent to inserting a zero-value pixel in-between each pair of original pixels and applying the full filter to this new image.