Derivative compressive sampling (DCS) is a signal reconstruction method from measurements of the spatial gradient with sub-Nyquist sampling rate. Applications of DCS include optical image reconstruction, photometric stereo, and shape-from-shading. In this work, we study the sensitivity of DCS with respect to algorithmic hyperparameters using a brute-force search algorithm. We perform experiments on a dataset of surface images and deduce guidelines for the user to setup values for the hyperparameters for improved signal recovery performance.