Some of the change-blindness demonstrations might fit the bill? Like the dalmatian example, most participants struggle to see the change item, but once they've seen it, as long as they remember that pair of images, by and large they identify the change item instantly. And it is possible to use much simpler stimuli than photographs.

Sony's website provides "a-ha picture"gallery by Dr. Kenichiro Mogi. Click each image, and you can see the original image by pressing a pink button left of the black-and-white image. Some of them are too easy but there are couple of images which are quite challenging.

http://msc.sony.jp/member/enjoy/ahap/ahapicture/backnumber.html

I guess you can make this type of images using image processing software pretty easily.

I am sure you would find that many of the illusions depending on the two dimensional imaging of a suggested there dimensional reality would be seen differently by those skilled in such matters.  I have lectured to designers and architects all my life and often early in the course discuss these problems for they rely heavily on being able to draw on a two dimensional surface a representation of a three dimensional object that so far only exists in their mind.

As Paul Knox has suggested change-blindness illusions work great in this regard. I also use a selection of dalmatian-like images/cutouts/shadow images that seem to do the trick. Always good to throw in an upside-down shadow image of a map among the usual faces!

Perception of ambiguous stimuli have the advantage, provided that you don't choose the most famous ones and that the presentation is short enough, to be stable as long as the participant doesn't know about the existence of the alternative interpretation. However, contrary to the Dalmatian illusion, one the percept become bistable, you can still measure the effect of prior knowledge. That is, if you modulate this prior via instructions, you will be able to measure the consequence on the static and dynamic parameters of bistability.

It is not facilitation, but it might help if you primary goal is to measure the influence of prior beliefs. 

This one is rather striking, even if it doesn't meet the specific requirements you mention. For one thing, it's auditory, rather than visual. But structurally it's quite similar to the Dalmatian dog example that you give. http://www.theatlantic.com/technology/archive/2014/06/sounds-you-cant-unhear/373036/  I am not sure what you have in mind by the "task" in the Dalmatian dog example, but if it's simply staring at the thing, trying (usually unsuccessfully) to see anything in  it, then there would be a comparable "task" in this auditory example. Cheers.

Have you thought about the hollow face mask illusion?

I endorse Sachi Mizobuchi's suggestion of using "Moody" images.

These have been used successfully in the past to probe effects of prior knowledge. See work by Sinha and Meng ;  Hsieh and Kanwiyher.

The coffer illusion may fit the bill. People tend to struggle to find the circles the first time. Then.. they can never be "unseen".

http://illusionoftheyear.com/2006/coffer-illusion/

This one should be a slam dunk. You see it immediately w/ prior knowledge and takes a long time to see it w/o prior knowledge. May not even need statistics to see the difference!..

http://www.michaelbach.de/ot/cog_dalmatian/

Superior mirage (very common in our island of Formentera). The mountain close-by the sun are Denia's 110km and Alcoy's 150km. If we can see them, it's because hot air got trapped under cold air. We also see Africa (200km), the other side of the island. We're used to it and join with friends to enjoy them. But tourists, rational ones, claim it's an illusion what they also see. Or an unknown island... ;-) They were feeling destabilized by this brutal change of surroundings, so they choose to totally ignore it.