It combines the squared log range with the squared log close-to-open move, weighting them to minimise variance under the assumption of a driftless continuous price process. The result is roughly seven to eight times more efficient than close-to-close-volatility.
That efficiency is bought with assumptions. It expects zero drift, so it underestimates volatility for a strongly trending instrument, and it treats the open-to-close and close-to-open periods with fixed weights that do not match every market's actual overnight-to-intraday split.
Use it where you have clean OHLC bars and a market that trades nearly continuously, such as index futures. For single stocks with heavy overnight news flow, prefer yang-zhang-volatility.
Related: parkinson-volatility, yang-zhang-volatility, ohlcv-bar, realised-volatility