Wavelet Transforms for Managing Landsat Data Sets on a Geospatial Client-Server Discrete Global Grid System

Landsat data sets are usually composed of a set of images at various spectral bands and different spatial resolutions. Since Landsat images are captured at different spectral bands, they may reveal important characteristics of a region. An alternative image is made to identify the characteristics of a region by a linear or non-linear combination of images at different bands. In this image, important features and characteristics of the region such as its vegetation and snow are identified. However, since Landsat data sets are of significant volume, visualizing these data sets is hard unless some types of compression or simplification are performed on these data sets. Many frameworks are available for working with geospatial visualization and analysis. Discrete Global Grid Systems (DGGS) are particularly important in this area as they provide a common framework to integrate and analyze a vast variety of geospatial data. DGGS benefits from a client-server system in which on the client side, users are manipulating, observing, and requesting data sets and on the server side necessary queries are addressed and required data sets are transmitted. As Landsat data set is large in its native resolution, working on and transmitting such large data sets might be inefficient. Besides, working with Landsat data sets on the client side might also be very time-consuming and inefficient as operations on Landsats are expensive in terms of time and space. We suggest a system to reduce the volume of the data and balance the cost of operations on the client and server side on a DGGS. To do so, we employ wavelet transforms. In wavelet transforms, a set of fine data sets F are decomposed into a set of detail vectors D, and coarse approximations C. Dimension of D and C together is equal to F and F can be reconstructed using D and C without losing any information. We provide the coarse approximation C of Landsat data sets on the client side. Users can easily work with the coarse version of the data set and ask for the reconstruction if they need a more accurate data set. Wavelet transforms can be performed on the images in different bands alongwith the temporal dimension. To address linear or non-linear operations, it is also possible to provide an approximation of the resulting image using C and then reconstruct the high-resolution image by sending appropriate details.