simply by looking into morphology is one way

Martensite has a much lowerlattice perfection so that this is a general approach. But in fact it is the same problem you have in XRD. It is quite difficult to uniquely differentiate between both signals. Moreover, preparation artefacts cannot be totally excluded so that the simple consideration of blurred pattern also caused by pattern overlapping cannot be the ideal answer. It has some reason, why the main tool for decades  is...the imaging in TEM and/or characterizing the morphology. There are several papers published over the last decade and indicate the application of EBSD for this task, but the missing general tool in software packages should tell you that the community does not doubt about the presented results but has doubt in a "blind" application of this tool in industry. Otherwise, the EBSD marked would explode because of the amazing interest of steel manufactureres. But martensite is a too complex material in order to press it in such a simple concept of EBSD pattern analysis as we do it now with the available software packages.

If you cannot resolve two phases (uniquely) with XRD, usually a differentiation with the lower resolving EBSD seems to be (from my impression in general) impossible. EBSD only has the advantage that you know where the signal has been formed which enables theoretically a characterization of boundaries. But if everything is highly (cold) deformed....what is your reference orientation for the misorientation analysis? You don't know what it going on slightly below the surface...

Ferrite being softer than martensite, in the EBSD maps of your cold deformed materials you can notice elongated grains (pertaining to ferrite)  in contrast to others structures/or substructures wherein subgrains should likely to be present due to high dislocations density (namely the martensite for example).

I would not argue with respected colleagues, however under certain conditions MRTNST is reliably discriminated from FRT by the GAM function (with gradations of gray or colors) in standard  EBSD soft :

1. tolerance for virtual "interfaces", as introduced for GAM, should be 5 deg and the scan step not bigger than 0.2 micron.

2. Cold deformation  should not be too much, ideally zero (after-quenching state). It is for M to persist much  more dislocated  than F. As far as disl densities of these species of alpha-phase initially differ by about two order, some 15....20% of cold strain seem to be still tolerable.

I imply that disl density in F will gradually grow, whereas the ultimate disl density of M will be almost invariant and "waiting" until reached by F.  

Many Thanks for sharing ideas

Martensite will have different morphology and it always possesses  crystallographic relationship with parenting austenite grain (KS or NW).

@ A.A. Zisman: You certainly know how GAM is calculated (I am not sure that the equation is always the same for all software packages), but do you really know, how the grain recognition based on EBSD data work, and how much this depends on the acquisition conditions, band detection, Hough resolution? Theoretically, I would agree with you, Alex, but as I mentioned above, it would be used if it would work in general and would be trustful.

The "free" parameter are simply (still) unpredictable, and the materials (undeformed or deformed, heat treated before transformation or not (residual Ferrite-Austenite orientation relationships) are too complex (my personal opinion). In your lab you can (perhaps) do this...

Why don't you use Kernel Average Misorientation (KAM) map in EBSD? KAM shows distribution of dislocations. Because transformed martensite contains lots of dislocations, I think that this technique can distinguish between transformed martensite and ferrite.

I performed a lot of maps with strain induced martensite in 304 SS. First of all you "grain" size of martensite at low strains can be around 50 nm and less so I suggest to use TKD (Transmission Kikuchi Diffraction). The easiest way to differentiate martensite and ferrite is to perform correlative EDX map of investigated region (I know it's not EBSD but it's most simple). The martensite will have the same chemical composition as austenite while ferrite will be enriched with Cr and Mo.

to all:

Do you know good reference with EBSD maps where strain induced martensite was analysed in duplex or austenitic stainless steels?

@ H.J. Sung: since KAM also only considers dislocations which ROTATE the crystal. The precision of EBSD measurements is not well defined, i.e. so your rotation. You even can have your crystal full of dislocations and you don't see any rotation, only a blurring of patterns (no KAM signal). In so far, KAM gives you in maximum a rough impression about for rotations geometrically necessary dislocations, but not about the global distribution of dislocations. Therefore, TEM is still much better, and even there you need to look from different directions in order to get an impression about the dislocation distribution of different character.