Habitat complexity is known to influence the structure of fish assemblages. A number of techniques have previously been used to measure complexity, including quantitative in situ methods, which can be time-consuming and labour-intensive, and more rapid semi-quantitative visual scoring methods. This study investigated the utility of a novel method for estimating complexity, whereby habitat height was measured using stereo-photogrammetry from diver-operated stereo-video, traditionally used to sample fish assemblages. This ‘stereo-height’ method was compared to established in situ and visual scoring techniques and found to produce similar estimates of complexity. To determine how relevant the proposed method is for assessing ecological relationships, it was then used in conjunction with visual scoring of relief and point-intercept samples of benthic composition to model fish–habitat associations in the Pilbara region of Western Australia. Visual scores of relief were marginally stronger predictors of fish assemblage parameters and functional groups than the stereo-height measurements, providing support for the visual scoring approach. The only exception was for corallivorous fishes, which were more strongly correlated with stereo-height measurements. This study has presented a method for assessing habitat complexity using video imagery that is both comparable to traditional in situ techniques and useful for investigating fish–habitat relationships. We suggest that future studies interested in collecting habitat complexity data from new or existing stereo-video samples use both the stereo-height and visual scoring methods presented here. Together these methods enable studies to rapidly and effectively assess fish–habitat relationships across a range of habitats without the need for in situ methods or solely relying on field observers trained in visual scoring techniques.