There are a number of things that determine quality, yet there are no hard and fast rules. However, I find simple images with no overlapping parts and few solid colors with a reasonable pixel count will convert well.
To begin, I look for the bitmap image file size to be around 200kb or larger. Smaller file sizes are tiny images at 72 dpi. If they are not simple, you are likely to get poor results. Solid colors that are not shades of the same color are easier to tell where red ends blue begins rather than red next to pink. Fades, shadows, and overlapping objects, especially objects of various shades of the same color, are more difficult for the software to tell apart. Shadows and fades are likely to be only partially picked up by your software, giving results that have lots of little objects you will likely want to delete.
—Bob Hagel, Eagle’s Mark