Usually minimaps are done by rendering the scene using a special, top-down camera. In your case though, the amount of scaling is extreme, so you'll lose a lot of detail one way or another. However, I suspect that your biggest issue is that the important fine detail is lost, like contours and edges.
I suggest you try using an image processing technique to preserve those edges. You can either find a library to do these, or if you know your way around bitmaps and multiplying pixels, they're fairly simple to implement too.
To demonstrate, I'll use this giant picture of the world map with political boundaries. Here's a close-up of the British Isles and an extremely scaled-down version (using basic bilinear, similar to what DirectX would be doing) to compare the loss of fine detail:
Resize and sharpen
One easy method is to perform gradual resizes, but also sharpen at each iteration to preserve the edge detail. In this example, I'll resize to 50% and sharpen the image repeatedly until the image is tiny. Here's the result of a few iterations of resize and sharpen:
And here's the final result, at about 258x128:
Another method is to use an edge detection image filter, thicken the edges, and combine the results with the normally-scaled image. Here's the result of one such filter:
Combined with the result of a normal scale, I get this: