Yes, trash can.
I'm not an engineer (I started out at Rutgers to be an EE but the draft got me), but maybe this explanation would help: If it worked for Bill Naivar at Georgia Tech, I don't see any reason why it wouldn't work for Trip. Did you look at the links that Trip previously posted?:www.prism.gatech.edu/~wn17/www.prism.gatech.edu/~wn17/Web%20bill%20page%202.htm
Bill grounded the can and used a feedthru for the coax so that there would be no chance of interference being picked up on the outside of the feedline.
Wouldn't the trash can scatter and reflect the signal into multipath oblivion?
On the contrary, the trash can narrows the beamwidth to ignore off-axis reflections and benefits from the internal reflections to enhance the direct signal; Bill mentions an increase in gain from the internal reflections that are close enough in time sequence so that the tuner can handle them. The ability of a tuner to handle multipath reflections is determined (in part) by the time difference between the arrival of the direct signal and the reflected signal. The greater the difference for acceptance, the better the tuner.
The British DVB-T system handles multipath better than 8VSB, but it requires a SNR greater than 21dB, which means more transmitter power for the same coverage area.
The term "anti-ghosting" is a holdover from analog days. We saw a secondary image on the screen from multipath reflections. The distance between the primary image and the ghost image is an indication of the time difference between the direct signal and the reflected signal.
The depth of the shielded enclosure is what narrows the acceptance angle. A yagi would require a deep enclosure, but a reflector type antenna doesn't need to be much than 1/2 wave deep (1/4 wave for the antenna to rear plus maybe another 1/4 in front....to be determined by experiment).
Using a long yagi for its narrower beamwith to fight harmful multipath reflections is a traditional approach, but using a "shrouded" antenna is even more effective.
Another technique previously used is to substitue a CM4228 for a CM4221 antenna for UHF multipath problems. The 4228 has a much narrower horizontal beamwidth because of horizontal stacking even though it doesn't have much more gain.
In Trip's location his local VHF signals are strong, so he doesn't need a high gain antenna, but he does need a very narrow beamwidth to reject off-axis multipath reflections, which is why I suggested a folded dipole or a fullwave loop as the active element. Either of these would be smaller than a fullwave bowtie so that the shroud could be smaller. A more expensive off-the-shelf solution would be an AD C5 in a can.