Reconstrucción de imágenes sparse de radar en alta resolución usando algoritmos SPGL1 y GPSR.

Abstract

In this paper grade new reconstruction techniques signals as are SPGL1 and GPSR, software level in radar images of high resolution, these images are in a (sparse) sparse domain is implemented, and otherwise the technique of the discrete cosine transform is used to convert these images not sparse to sparse, this is to have a successful reconstruction. Then it is shown that the theory actually Compressive Sensing (CS) can be applied to image reconstruction of high-resolution radar, as well suggests the theory. The radar image in sparse state, will be screened on a random matrix, which will compress the image to 10%, 25%, 50% and 75% this process will allow us to see the reconstruction of the image in each of these compressions and also it analyzed with a comparative study the performance of each reconstruction methods, using different criteria such as loss of data and computational cost in order to make known which of these methods is the most efficient for each type of application. This work jointly presents the simulation of these reconstruction techniques in MATLABlevel code, known as guide tool will be a graphical user interface, additionally we will

have a user manual which will consist of the steps to experience reconstruction of an

image.

The

methods were effective in reconstructing radar images in high definition, was presented loss of data or failure of different reconstruction for each method, SPGL1 obtained errors for image 1 to different degree of compression (10% = 0.2311, 25% = 0.8245 50% = 2.7630, 75% = 7.3135), for GPSR was obtained (10% = 5.2220, 25% = 6.8218, 50% = 7.3108, 75% = 9.7869) solver SPGL1 proved best in terms of quality of reconstruction.