The Hubble constant ($H_0$) measures the current expansion rate of the Universe, and plays a fundamental role in cosmology. Tremendous effort has been dedicated over the past decades to measure $H_0$. Notably, Planck cosmic microwave background (CMB) and the local Cepheid-supernovae distance ladder measurements determine $H_0$ with a precision of $sim 1%$ and $sim 2%$ respectively. A $3$-$sigma$ level of discrepancy exists between the two measurements, for reasons that have yet to be understood. Gravitational wave (GW) sources accompanied by electromagnetic (EM) counterparts offer a completely independent standard siren (the GW analogue of an astronomical standard candle) measurement of $H_0$, as demonstrated following the discovery of the neutron star merger, GW170817. This measurement does not assume a cosmological model and is independent of a cosmic distance ladder. The first joint analysis of the GW signal from GW170817 and its EM localization led to a measurement of $H_0=74^{+16}_{-8}$ km/s/Mpc (median and symmetric $68%$ credible interval). In this analysis, the degeneracy in the GW signal between the source distance and the weakly constrained viewing angle dominated the $H_0$ measurement uncertainty. Recently, Mooley et al. (2018) obtained tight constraints on the viewing angle using high angular resolution imaging of the radio counterpart of GW170817. Here we obtain a significantly improved measurement $H_0=68.9^{+4.7}_{-4.6}$ km/s/Mpc by using these new radio observations, combined with the previous GW and EM data. We estimate that 15 more localized GW170817-like events (comparable signal-to-noise ratio, favorable orientation), having radio images and light curve data, will potentially bring resolution to the tension between the Planck and Cepheid-supernova measurements, as compared to 50-100 GW events without such data.