One does not simply take a very good looking "big" multi-coloured image and shrink it down to about 48 per 48 pixels to obtain a very good looking "small" image. The likely result is a bad looking image.
The way the eyesight works is that the colour from a point is decreasingly spread over nearby visual areas in our perceived image, so images using few pixels have each pixel dirtying with its colour all nearby pixels, whereas such effect is virtually irrelevant for big images.