Controlling the interface between Bitumen-froth and Middlings in separation cells in the oil sands industry is important for economical and environmental reasons. Traditional sensors do not provide reliable measurements of this interface level and image based sensors are being used to alleviate this problem. Previous work in this area has focussed on separation cells with a single side-view glass. The current work describes a new image based algorithm for interface level detection and confidence estimation based on the concept of image differencing. The algorithm can be extended in a straight-forward manner to separation cells with arbitrary number of side-view glasses. Off-line and on-line results show that the algorithm accurately detects the interface level in normal process conditions and outputs correct confidence values in other situations with very low false positive and negative rates.