The difference is more to the subject of the database than anything else. By the way, Inmon approach focus on normalizing as much as possible to make the ETL process easier and less error-prone; but non-normalized data are acceptable on data marts, even to Inmon.
OLTP: it is a full normalized database (commonly on 3NF) focused on a specific application. Usually, the data changes are small and happen on the row level.
Inmon DW: this is the central information database for the whole organization. The modeling is focused on business needs and this database centralize all the information from every system in your organization. IE: you can have Customer information in many of your systems, where each system brings you a complementary information. Following the Inmon approach, you have to have only one Customer entity on the DW, and this entity will have all the information from every system, in a conformed way (so it is not only copying tables, you should efficiently relate data from different systems inside your DW). Comparing with the Kimball approach, where classifications (hierarchies) can be simple columns in a dimension table, Inmon would create separate tables to "normalize" this info and make the ETL process easier.