[robski]

There is an element of truth in what Foxy is saying if the original image is poor. i.e blurred. Generally the image will look sharper when down-sampled.

If the original is sharp then you are at the mercy of the re-sampling algorithm. Some are quick and dirty while others try to take into account what else maybe happening in that area of the image which entails more processing.

I am often accused of pixel peeping but I find it helps explain what is really happening. I've knock up a couple of simple images in Photoshop and down sampled them using a crude method and the bilinear method to compare against the original. The set in the top row are of a diagonal line with random noise added. The bottom set are just the line. The original was 480 x 480 pixels and down sampled to 233 x 233 pixels. The attached are 50 x 50 pixel crops blown up by 600%. In the crude method the resultant pixel is simply voted in or out by the majority. The bilinear method has dithered to retain some data instead of completely removing it. The dithering has caused a blurring hence the need for some degree of sharpening after the re-sampling.

The method outlined by Mark is tried and tested and gives good results.