Yes, two-state "dim" and "bright" is easy.
Given any circuit that blinks a LED hard on and off with a transistor, you add one resistor across that transistor. Then when the transistor is fully off, the resistor allows a dim glow. I would start with a resistor of exactly the same value as the current limiting resistor already connected to the LED.
(Every LED needs a current-limiting resistor).
For a few markers, a independent battery and 555 timer at each one is going to be the simplest marker hardware. (plus a few resistors and capacitors).
The overall system simpler if you can synchronize the LEDs: turn all markers on at the start of the cycle, then turn off one marker at a time until they are all off, then turn them all back on and start the cycle over. The amount of energy needed to keep a bunch of LEDs blinking for a couple of hours typically weighs much less in the form of one or two central batteries rather than one battery per LED.
(This requires a comparator IC at each LED, or a few shift registers or an Arduino emulating those shift registers at some central location).
(This requires lots of wires stringing from one marker to the next, or from each marker to some central point -- so that may not be possible for your application.)
It makes your vision recognition software much simpler if the PC can control the LEDs directly. Then when the PC is searching for LED_5, it can turn off and on LED_5 and be confident that the one LED that blinked must be LED_5.
Perhaps using something like a USB to 8 bit parallel port converter, which (with 8 resistors, one per LED) can directly control 8 LEDs or (with 4 resistors, one per column) a 4x4 matrix of 16 LEDs.
(This requires yet another wire, a USB cable from the PC to the converter, but it doesn't require any batteries or transistors or additional chips -- this may be the simplest for a programmer non-electronics guy to get working).