Several things have to be done here.
First, deinterlace the video with a good method. You cannot resize an interlaced video very well. Using
VirtualDub, two easy-but decent options exist: Yadif (built into
VirtualDub 1.9.x), or Deinterlace Area-Based. While some purest would insist Yadif
must look better, it's not always the case -- try both. I used Area-Based on the test clip.
Next, you have to upscale the image, by cropping off the sides. Remember that your input is 720x480. That's a ratio of 3:2 (240x3=720 x 240x2=480), so we want to cut off 3x2. For example, 15 pixels from left/right, 10 from top/bottom. That won't be enough here, but you get the idea now. For the image you showed, I used 100+66 (50 L/R, 33 T/B). In VirtualDub, you pick the resize filter, resize to 720x480, and the go to the cropping options and cut off your pixels.
So now your image is dominant on-screen, not a small distant-looking box .
If you want to further mask that overcan to a clean box-like shape, you again run a resize filter, but this time you pick the letterbox/masking option, and set it to 720x480. Then enter the cropping and chop pixels off all sides. 3:2 is suggested, but not required. For TV overscans, the left/right loses more than top/bottom. About 16 off L/R (16L,16R) and 8 off T/B (8T,8B) is good. For your example, I think I did 22 L/R and 12 T/B -- and it should still be fine on a TV, not show borders. Note that preview doesn't always work here, just wait for the "full preview" in the output window of VrtualDub. VirtualDub has a bug with plugin-view preview (in crop mode), when you stack resize filters.
Remember that your video is now progressive. I suggest leaving it as such, don't re-encode it to bottom-field MPEG-2.
The VCF file, sample MPEG, screen shots of working in VirtualDub -- that will come soon. Just can't get to it right now, wanted to give you the text-only version for now.