Electrical capacitance tomography (ECT) provides a means for non-invasively imaging multiphase flows, such as those in fluidised beds. Traditionally ECT images are reconstructed using the assumption that the distribution of permittivity varies smoothly throughout the sensor region. However, for many applications there are step changes in the permittivity, for example, between the bubble and particulate phases in a fluidised bed, and the assumption of smoothness is flawed. In this article a Total Variation Iterative Soft Thresholding (TV-IST) algorithm is used to reconstruct ECT images that allows for sharp transitions in the permittivity distribution. This new algorithm has been compared with established algorithms for ECT image reconstruction. It was found that the TV-IST algorithm reduced the sensitivity to the threshold level chosen when extracting measurements of bubble size from ECT data sets. Measurements of the bubble size distribution in the fluidised bed using the TV-IST algorithm agreed closely with established empirical correlations for the size of bubbles. The results demonstrate that ECT can provide accurate and high spatial resolution measurements of features such as bubbles in gas-solid fluidised beds.