The Hubble Space Telescope (HST) has a point-spread-function FWHM of 0. With this incredible spatial resolution, it is now possible to classify high-redshift galaxies according to the Hubble sequence, and, to quantify galaxy morphology with parameters such as central surface brightness, axial ratio, bulge-to-disk ratio, disk scale length and light profile models (point source, r
, exponential). Although samples are still fairly small, very interesting results have been obtained, and HST imaging offers the exciting possibility of determining the morphology of the galaxies responsible for the faint galaxy excess.
Axial ratio is the simplest morphological parameter to measure. It appears that the axial ratio distribution of small (half-light radius < 0), faint (I 
20.5) galaxies with exponential surface brightness profile has a marked excess at ratios around 0.7 (round = 1.0) over local samples of spiral galaxies [\protect\astronciteIm et al.1995]. The excess galaxies are similar to local dwarf galaxies: same axial ratio distribution and colors ((B
V)
0.4, (UB) 0.2). These ``Small Exponential Elliptical (SEE)'' galaxies make up 2125
of the population mix at 20 < I < 21. Combined with irregular/peculiar galaxies, they could be responsible for up to 80 of the galaxy excess over model predictions. The presence of such large numbers of SEEs could be explained by a steep (
1.4 to 1.8) local luminosity function or a starburst stage around z0.5 caused by minor mergers.
Other studies [\protect\astronciteGlazebrook et al.1995b][\protect\astronciteDriver et al.1995] have produced deep (I 24.2 or B 26) morphological number counts based on light profiles, bulge-to-disk ratios and direct images. The counts showed that the galaxy mix at faint magnitudes differs from our local neighborhood. There was a steep rise in the number of late-type and irregular/peculiar galaxies. Sd/Irr galaxies made up 3050 of the total population as opposed to 810 in the CfA survey [\protect\astronciteMarzke et al.1994a]. The number counts for early-types (E/S0, Sabc) were consistent with little or no evolution if number count models were normalized to the observed counts at b
= 1820. This normalization was twice as high as the 
value derived for local surveys [\protect\astronciteLin et al.1996][\protect\astronciteLoveday et al.1992]. The number counts of late-type/irregular galaxies were modelled in four different ways: (1) a no-evolution model based on the Loveday luminosity function for late-types, (2) a no-evolution model based on the Marzke LF [\protect\astronciteMarzke et al.1994a] for late-types, (3) an evolving model with a burst in the Sd/Irr population at z = 0.5, and (4) a dwarf-rich model with 
= 1.8 and a free normalization chosen to fit the counts. Models (1) and (2) under-predicted the number of observed late-types/Irrs. In model (3), the increase in luminosity required for the entire late-type population to match the counts with the Loveday et al. loveday92 LF and with the Marzke et al. marzke94b LF were 
= 2.0 magnitudes and = 1.3 magnitudes respectively. The counts could also be matched with the Marzke et al. marzke94b LF and a = 2.0 mag increase in luminosity in 20 of the late-type population. These increases in luminosity are consistent with a 1 Gyr burst in a dwarf galaxy followed by an exponential fall-off (see Section 
). In model (4), a value of of 
Mpc
was required to match the counts. This is five times the Loveday et al. loveday92 normalization, and it is inconsistent with faint redshift surveys as it predicts too many low-redshift objects.
Three interesting results came from a quantitive study of a sample of 32 galaxies with magnitudes 17.5 < I
< 22.5 and redshifts 0.5 < z < 1.2 [\protect\astronciteSchade et al.1995]. First, galaxies at z 0.75 exhibit the same range of morphological types as seen locally (ellipticals, spirals and irregulars). Second, 30 of the sample were so-called ``blue nucleated galaxies''. They had asymmetric/peculiar structures and blue, compact components not always centered on the galaxies. Diagnostic line ratios ([OII]/H
and [OIII]/H) indicated that the compact components were sites of star formation. As shown later (section ), these blue nucleated galaxies appear to be directly linked to the peculiar [OII] kinematics observed in our survey even though it covers a lower range of redshifts. Third, the central surface brightness of galaxies at z 0.7 was 
= 20.2 
0.25 mag arcsec
i.e. 1.2 mag brighter than the Freeman value found in local spiral galaxies. This increase in rest-frame surface brightness indicates that galactic disks undergo strong evolution at those redshifts, probably as a result of global star formation.
The work of Schade et al. schade95 has recently been extended with dramatic results [\protect\astronciteSchade et al.1996b][\protect\astronciteSchade et al.1996a]. They studied the surface brightness of 351 cluster and field disk galaxies over the redshift range 0.1 < z < 0.6, and 166 cluster and field early-type galaxies. Both samples were drawn from the CNOC galaxy cluster survey [\protect\astronciteYee et al.1996][\protect\astronciteCarlberg et al.1994]. Disk and early-type galaxies evolve significantly over that redshift range. Moreover, there was no significant difference in evolution in clusters and in the field. At redshifts of (0.23,0.43,0.55), the disk surface brightness in cluster and field late-type galaxies was higher in the B-band by 
(B) = (0.580.12,1.220.17,0.970.2) mag respectively compared to the local Freeman law. For early-type galaxies, their surface brightness increased by (0.250.10,0.550.12, 0.740.21) mag at redshifts of (0.23,0.43,0.55) compared to a local z=0.06 M(B) versus R
relation. The amount of brightening was consistent with passive evolution of an old, single-burst population. The fact that galaxies evolved similarly in clusters and in the field was remarkable.
HST imaging has shown that late-type/irregular galaxies were responsible for the faint galaxy excess. The exact amount of luminosity evolution remains dependent upon uncertainties in the local luminosity function, but plausible models indicate that it must be at least one magnitude by z = 0.5.