Performance Analysis of CS Method for Image Reconstruction on Microwave Tomography

Abstract

This paper describes the performance analysis of image reconstruction on microwave tomography using the compressive sensing (CS) method. The investigation is carried out by processing an image of tree trunk object with different pixel sizes, i.e. 256×256, 128×128, and 64×64. Performance comparison are assessed using mean square error (MSE) and structural similarity index measure (SSIM). The results of image reconstruction using CS method with 256×256 pixels have the best MSE and SSIM with values of 0.001 and 0.944, respectively. Meanwhile, the image reconstruction with 128×128 pixels is not much different from the one with 256×256 pixels which produces the MSE value of 0.001 and the SSIM value of 0.899. Furthermore, the result of image reconstruction with 64×64 pixels is worst among others and yields a blurry image. However, it has MSE and SSIM values of 0.001 and 0.837, respectively, with fastest computation time compared to other pixel sizes.