The on-line images of Codex Bezae (mentioned already here) are very good. One of the particular difficulties in dealing with and imaging Bezae is the extreme thinness of the parchment (mentioned briefly in Parker, Codex Bezae, 22f). This means that quite often bleed-through from the other side of a page makes reading difficult, and checking readings in the manuscript itself often involves checking both sides of the leaf. The new images (unlike the old facsimile) seem to have been taken in such a manner that minimises this bleed-through from the other side (although I couldn't find any discussion of imaging issues on the web-site and we should note that imaging always involves some level of interpretation/manipulation of the data, see here for an interesting recent discussion).
Anyway, here is an example from Matt 2.21. In the old facsimile it looked like this:
The end of the second line was read by Scrivener as EIS THN ISRAEL (Swanson also thought D* had THN, corrected to GHN). I have suspected for a while that this (which would in any case have been an odd reading) was wrong, and the new images (image 14, 6v) confirm this:
The parallel to the left of the upright is from the other side. Bezae has a normal gamma, EIS GHN ISRAEL (one can also check this by comparing all the gammas and taus on this page).