The empirical wavelet transform is a data-driven time-scale representation consisting of adaptive filters. Its robustness to data has made it the subject of intense developments and an increasing number of applications in the last decade. However, it has been mostly studied theoretically for signals so far and its extension to images is limited to a particular mother wavelet. This work presents a general framework for multidimensional empirical wavelet transform from any mother wavelet. In addition, it provides conditions to build wavelet frames for both continuous and discrete transforms.