Aaron,

You can cancel flicker noise if you are doing a sampled-data peak detector.  If you take two samples in a row, and subtract them, then noise that is much lower in frequency than your clock rate will appear on both samples and be cancelled.  This is called Correlated Double-Sampling.

In a continuous-time system, you can do a kind of dual-slope peak detector.  This is similar to a dual-slope ADC.  Essentially, you integrate the signal for some set period of time in one direction, then you integrate a reset signal in the other direction for the same amount of time.  The idea is that the flicker noise will be the same on the two integrations and then cancel.  For this you need an accurate time base and constant current source.

Most likely, this is overkill.  You may just need to size your devices appropriately to get the flicker noise below your noise floor.  It really depends on the accuracy you need for your application.

Carl