Classification of seafloor habitats from geomorphological proxies is increasingly being applied to understand the distribution of benthic biota, particularly as more and larger datasets, collected with high-resolution multibeam echo sounders, are becoming available. With increased capacity to collect and use sonar data, there is a need for automated approaches to identify seafloor structures and habitats. For a survey area of approximately 2.5 × 2.5 km, a generic feature extraction algorithm similar to terrestrial topographic analysis has been developed and applied to a spatially complex submerged mid-shelf reef in the Great Barrier Reef lagoon. Multibeam data collected with a RESON Seabat 8101 was gridded at 1 m resolution, and an automated feature extraction method was applied that analyses the seafloor geomorphology to predict reef features from geomorphological proxies. Quadric surface fitting was used to determine various surface parameters based on multiple spatial scales. Subsequently, 6 morphometric feature types (plane, channel, ridge, pass, pit, and peak) were derived for all mapping scales. Weighted multiscale fuzziness was then applied to extract the dominant morphometric feature classes. Finally, seafloor morphology patterns in combination with seafloor rugosity measurements were analysed in order to predict reef features. These reef features cover significant patches of bioturbation beyond the submerged platform reef, as well as the dominant reef features such as outer-reef crest and inner-reef flat. Layback-corrected and manually classified towed video transects support the classification algorithms used to extract the reef features.