We present the analysis of deep colour-magnitude diagrams (CMDs) of 6 stellar fields in the LMC. The data were obtained using HST/WFPC2 in the I and V filters, reaching V=26.5. We discuss and apply a method of correcting CMDs for photometric incompleteness. A method to generate artificial CMDs based on a model star formation history is also developed. This method incorporates photometric error effects, unresolved binaries, reddening and allows use of different forms of the initial mass function and of the SFH itself. We use the Partial Models Method, as presented by Gallart and others, for CMD modelling, and include control experiments to prove its validity in a search for constraints on the Large Magellanic Cloud star formation history in different regions. Reliable star formation histories for each field are recovered by this method. In all fields, a gap in star formation with t=700 Myrs is observed. Field-to-field variations have also been observed. The two fields near the LMC bar present some significant star forming events, having formed both young (t < 1Gyr) and old (t > 10 Gyrs) stars, with a clear gap from 3-6 Gyrs. Two other fields display quite similar SFHs, with increased star formation having taken place at t=2-3 Gyrs and 6 < t < 10 Gyrs. The remaining two fields present star formation histories closer to uniform, with no clear event of enhanced star formation.

We analyse a sample of over 1000 stars from 32 fields imaged in the V and I bands with the Wide Field and Planetary Camera, on board of the Hubble Space Telescope. The fields are located at Galactic latitudes | b | >= 15deg and in various directions on the sky. We consider models for the structure of the Galaxy with different choices for the main parameters governing the shape and luminosity function of the thick disk and stellar halo. Comparing model predictions with the observed colour-magnitude diagram we are able to rule out an increasing or flat stellar luminosity function at the low-luminosity end. We also rule out large values of the vertical scale height of the thick disc, z_0, finding it to be in the range 800 <= z_0 <= 1200 pc. As for the local density normalization, values within the range 4 % <= n_0 <= 8 % seem to better reproduce the data. Our data essentially rule out a flattened stellar halo (c/a <~ 0.5) or models with both large local normalization and effective radii.