These are HST ACS/SBC F150LP far-UV (~1610 Angstrom) images of the HDF-North and the UDF 
outlined in Teplitz et al. (2006) and Siana et al. (2007), respectively.  Due to varying 
coverage across the images, the depths vary, but the majority of the area is deeper 
than m_AB = 28.0 (3sigma, 1" diameter aperture, See Figure 6 in Siana et al. 2007).  
Both fields (HDF-N and UDF) have three released images, listed below. 

HDF-North

hdf_f150lp.fits				Science image
hdf_f150lp_wgt.fits			Weight map
hdf_f150lp_exptime.fits		Exposure time map

UDF

udf_f150lp.fits				Science image
udf_f150lp_wgt.fits			Weight map
udf_f150lp_exptime.fits		Exposure time map

The science images are in counts/sec and the zeropoint is 22.45 (AB).  Thus, to convert a 
total flux (in counts/sec) to a magnitude, m_150lp(AB) = 22.45-2.5*log(flux). 

The weight map can be used to get the error based on the background (dark current).  The error 
(in counts/sec) at each pixel is 1/sqrt(wgt).  To get an error for the entire aperture, simply 
add the individual pixel errors in quadrature.

Using the weight map will only give an error based on the background, and will not include the 
Poisson error from the flux of the detected objects.  If you need Poisson errors on objects 
detected at very high significance, than you'll have to convert the object flux (in counts/sec) 
to total counts using the exposure time map (units in seconds).  With this, the total counts 
can be computed to determine a Poisson error.