Abstract: We propose extension of optical flow constraint (OFC) based method to third dimension (3-D) to solve the problem of estimating gas bubbles in a melted glass. We use a single camera whose distance to the fluid is variable in time and is used to capture a sequence of frames at different depths. Common 3-D techniques may not be applied to such cases where objects are not static and hence two frames at different heights at the same time may not be available for analysis. Based on sparse information available, we estimate different parameters around a given pixel using a robust estimator. We compare our approach to existing benchmarks in 2-D case using a practical application and evaluate it to a synthetic problem.