Processing of fused optical satellite images through parallel processing techniques in multi GPU

Technology makes many of the tasks that were previously difficult to perform, nowadays can be solved, one of them is to be able to carry out studies on large tracts of land and at the same time be able to have a level of detail of them, through the study of the satellite images provided by the Earth observation satellites, these images are composed of a series of spectral bands that will depend on the type of satellite mission that was conceived and the optical instrument that is found as a payload, these images are represented by multidimensional arrays and large size, so computational high computation equipment is required to process the images, added to this requires specialized software that allows the visual interpretation of satellite images. To be able to work with satellite images you have many configurations, normally you work with the configuration of separate bands that consists of working separately with each band of the image, these images have a particularity, the high resolution image is the one that is found in the Panchromatic band, where the maximum spatial resolution of the image is presented, which can range from metric to sub-metric, then the red, green, blue, and shortwave and wave infrared bands mean, these bands are in a lower range of spatial resolution for example if a satellite has a spatial resolution of 1 meter in the panchromatic and has 4 spectral bands (Red, Green, Blue, Near Infrared), these will have the resolution of 4 meters, so the level of detail is lost compared to panchromatic images. In order to improve this image performance we have the image configuration fused, where the resolution of all the bands including the color and infrared have the resolution of the panchromatic, this means that all are of resolution 1 meter, gaining resolution spatial and also this new configuration of the image has the color of the bands gaining spectral resolution, therefore these images have a greater weight in GB and its matrix increases in size, therefore it requires more processing time of the same, In the present article we present a technique that improves the processing time of the fused satellite images using parallel processing by using two graphic processors, with this the image processing task is distributed, as Matlab software was used as tool, because it allows us to manage multidimensional matrices and also allows us to der to have access to the graphic processor, we worked with 2 cards model GTX1050Ti of Nvidia.