Geographia Technica, Vol 13, Issue no. 1/2018, pp. 20-29


Marilyn CEPEDA, Iván PALACIOS, Alfonso TIERRA, Eduardo KIRBY

DOI: 10.21163/GT_2018.131.03

ABSTRACT: Multispectral satellite images are tools that allow the analysis of phenomena developed on the Earth's surface without being in contact. It is a raster model so it is possible to decompose it into a digital signal. There is a certain data that presents alterations (noise) due to errors caused by the sensors, atmospheric conditions, among others. Such examples affect its use and its derived products. Satellite images by their nature present difficulty in their processing and handling due to the considerable weight they have; whose problem justified the present work. The objective is to minimize white noise and to compress the image with the least possible loss of information through the Multiresolution Analysis (MRA) technique and Wavelet transformation. The images worked belong to the National Recreation Area "El Boliche" (Ecuador) that is next to the Cotopaxi volcano. Through a standard deviation evaluation of the obtained wavelet coefficients, the order of the "Discrete Wavelet Transform" (DWT) was established in the Daubechies (db) and Haar families. With db3 level 4, obtained a compression of 11.268% in respect to the original weight and with Haar level 4 11.288% as the best results. The wavelet db is more effective than the Haar type for the treatment of multispectral satellite images in the elimination of white noise and compression by means of the MRA, with a reconstruction of the signal without loss of information due to the type of wavelet used, which is evidenced in the image.

Keywords: wavelets, MRA, satellite image, noise, compression.

Full article here