Accurate prediction of wavelet coefficients relies on an understanding of the phase effects of edge alignment. This research examines techniques for uncovering edge information based on the available coefficients. These techniques are evaluated in the context of reconstructing an image from quantized wavelet coefficients. A predictor is described which can be trained on the coefficients to capture relationships among the pixels. Another method is presented where the quantized coefficients are interpolated to recreate an underlying continuous function. Based on this research, the interpolation process offers more promise in solving the dequantization problem.