The paper provides an evaluation of the usefulness of wavelet image compression for Internet GIS viewers. The most popular method of lossy image compression is called JPEG. This method is very often used in Internet both for images of natural scenes and for other types of graphics like maps. The latest research has proved that wavelet compression is a much better solution than JPEG method. Wavelet transformation comes from the Fourier family, but it describes a frequency function.like Fourier transform as well as a spatial function. Thanks to this we do not have to divide the image into small parts in order to realize local transformation as in JPEG. Like the JPEG the wavelet compression is most efficient for multitonal images. Results of application of the wavelet compression to images such as maps were presented in this study. The Mallat solution to wavelets representation of images was used. The image is decomposed into sets of four images with two times smaller resolution. This task is solved by means of simple lowpass and highpass filters. Lowpass filtering first along rows and than along columns of the image with a double reduction of size creates a medium image - LL (low-low). The next two images are the results of lowpass and highpass filtration LH (low-high). The last component is a result of highpass filtration applied in rows and columns - HH (high-high). A part of typical topographic map was used in the research. The map showed on fig. 1 has included eight colors. First the map was transformed from RGB color model into IHS model. The next actions were executed only for intensity channel the map was treated as a gray scale image. For the research needs there were two versions prepared from the original map. The first version of map contains eight intensity values as follow: 124,125,131. The second map.s version contains eight intensity values 80, 90, 150. Both images were transformed by using Haar method, which is the most simple wavelet transformation. Fig 2 shows four subimages (LL, LH, HL, HH) which look like both variants of map. The difference lies only in another range of value. Then, the Haar transform for both variants of maps was filtered. As the filtration results all values in the three detail subimages (LH,HL,HH) were reset to null. The result of the filtration is shown on fig. 3. Now each of these transforms of two map.s versions was reconstructed by using reverse Haar transformation. The result of reconstruction of both versions can be seen on fig.4 in comparison with the original map. The empirical research proves that the scaling of intensity value influences the reconstruction result. To recapitulate, the wavelets compression for images with a few tones should be preceded by data preparing in order to localize the index-data in a small sub-range within full available range (usually 0-255 i.e. one byte per pixel). Present users of internet expect that GIS-viewers will show high quality maps with high speed. The wavelets compression is a good tool for this purpose but must by used deliberately.