This paper analyzes images obtained from an automatic total sky imager 4. A preliminary classification of optical structures observed on images is proposed in RGB color space of attributes. Pixel RGB decomposition is done using a GDI 5 tool programmed in C++ language. This paper shows that RGB space could be used to observe the selected structures, group then in patterns and clearly find their occurrence locus in space. The objective of this study is to improve the automatic identification of patterns of scattering and absorption present in the atmosphere, captured by digital cameras from the surface for multiple remote sensing analyses. Palavras-chave: image processing, atmospheric patterns, computer vision, absorption and scattering, processamento de imagens, padrões atmosféricos, visão computacional, absorção e espalhamento.