What is it then?
Pixelfusion is basically an image resampling method. There are many resampling methods already available, ranging from point sampling, to bilinear, bicubic or even windowed sinc. Pixelfusion is different to all of these in one very important aspect, as we will find out later.
So what exactly is resampling, and why do we need it?
A single frame in a video, or image, is made of a grid of different coloured dots, or 'pixels'. The size of the grid is described as the number of columns x the number of rows of pixels, or width x height, and this is usually referred to as the resolution. An example of the resolution of an image might be 512 x 384, where it would have 512 pixels going across, and 384 pixels going down.
The display on a computer also has a resolution, which defines the number of pixels that are actually displayed on the screen. This is typically 1024 x 768. An image is displayed by setting the display pixels to the same colour as the pixels in the image. In our example, the image does not have a sufficient number pixels to fill the entire display, it will in fact fill exactly a quarter of the display.
This would be fine if the image was displayed in a window, but if the image was required to fill the entire screen there must be a method for increasing the resolution of the image so that it matches the resolution of the display. The simplest method to do this is to simply duplicate the pixels, this is point sampling. In our example, each pixel would be duplicated to become a group of 4 pixels next to each other, arranged as a square of 2 x 2. Put these groups of pixels together, and you eventually end up exactly filling the screen.
Point sampling is great, the only problem is that it doesn't look natural. The results look pixellated, this is great as an effect in a paint package but not when watching a video or viewing an image. The technical term for the error produced is aliasing. At the boundary of each group of pixels there are sharp edges, which do not exist in the original image. The accepted method for removing the fake edges is to perform a filter, to literally filter out the bad edges. Bilinear and bicubic filtering are the industry standard algorithms, bicubic generally giving the smoothest and more natural results.
After bilinear/bicubic filtering, the image now looks natural and faithful to the original, but it appears blurry. The filtering has removed the high frequency components of the image, or in other words, the 'crispness' of the image. The image can be sharpened to some degree, but this can result in undesired halo effects around edges, more to the point it still does nothing about the lost frequency components.
Pixelfusion brings back definition to video and images.
During the Pixelfusion process, parts of the image where there should be high frequency components have a special algorithm applied to them to actually add details back into the image. The result is a much clearer image, while still remaining natural to the eye.
We have developed 5 enhancement modes, each of varying complexity and quality :- 4 modes are designed for realtime video (low, medium, high and ultra), and one mode is designed for images. The current mode can either be selected manually, or automatically. The automatic mode uses factors such as the current CPU load, image resolution and video framerate to determine the highest quality Pixelfusion mode which can be used without causing the video to skip.
Each mode is fully optimised in assembler, using Intel MMX or SSE2 instructions if the CPU supports them, for the fastest execution possible.
Experience neuview for $12.99