It is common to find that for high frequency data, the seasonal adjustment procedures are linked to Model Based Approaches (MBA), like using GARCH and ARIMAs. Some time ago I found a paper about hourly adjustment using machine learning maybe if you can find it, it would be useful. And the following paper about why not to use H-P filter.
I think Fourier series is the best to de-trending the seasonality in the time series. You may find that in text books on Fourier series. Just search on the web you will find plenty of material on this issue.
The Dynamic Harmonic Regression (DHR) algorithm in the CAPTAIN Toolbox for Matlab will do this for you and is much more flexible than the Hodrick-Prescott filter (see attached paper). DHR has had considerable application in environmental data analysis over many years. The CAPTAIN Toolbox can be downloaded via a link at my CAPTAIN Web-site: http://captaintoolbox.co.uk/Captain_Toolbox.html/Captain_Toolbox.html
Other publications on this subject, including the main paper on DHR ( P. C. Young, D. J. Pedregal, and W. Tych. Dynamic harmonic regression. Jnl. of Forecasting, 18:369–394, 1999.) can be obtained from me via email at [email protected].