Reduction of random noise from multiband image data using phase relationships among their Fourier coefficients

• Published in: IBM journal of research and development 1984-07, Vol.28 (4), p.399-411
• Main Authors: DAVE, J. V, GAZDAG, J
• Format: Article
• Language: English
• Subjects: Exact sciences and technology; Fundamental areas of phenomenology (including applications); Imaging and optical processing; Optics; Physics
• ISSN: 0018-8646; 2151-8556; 0018-8646
• DOI: 10.1147/rd.284.0399

In connection with geological and geophysical explorations, multiband digital image data of the terrestrial surfaces are routinely acquired with the aid of multispectral scanners aboard aircraft or advanced spacecraft. The image data contain some noise due to various factors such as platform vibration, scene variability, atmospheric inhomogeneity, and digitization. The present report is concerned with the problem of the reduction of random noise from the multispectral image-line data by transforming them from the wavelength-space domain to the wavelength-frequency domain, taking into account the utilization of Fourier transformations. Attention is given to the filtering of multichannel data using coherency measures, aspects of multispectral classification, basic data sets, phase relationships, coherency measure variations, noise reduction for the single-line cases and for the training sites, and multispectral classification results.