This study describes the development of a similarity index for comparing the quality of a lossy compressed spectral image. The index is based on the structural errors that are computed between the original spectral image and the spectral image compressed in a lossy manner. The index is an average of the local indices obtained from using a three-dimensional sliding window over the images. We compare the performance of the new index to the two-dimensional index. The main advantage of the new index is that it can detect both structural changes from spatial domain and color changes from spectral domain. It has several parameters and they can be used to adjust the sensitivity of the index.