Target Detection of SAR Image Based on Wavelet and Empirical Mode Decomposition

Abstract

Target detection is a hot issue in synthetic aperture radar (SAR) image applications. The complex imaging mechanism of SAR imaging leads to low signal to noise ratio (SNR) of SAR images, which brings great challenges and difficulties for target detection. Therefore, the primary task of improving the target detection rate of SAR images is to increase its SNR, namely, to expand the difference between the gray value of the target and the background area. The coherent imaging mechanism makes a lot of speckle noise in SAR images, bringing a great impact on target detection. To restrain the speckle noise and improve the SNR, this paper proposed a new SAR image target detection method called the SWT-BEMD algorithm, based on the two-dimensional stationary wavelet transform (SWT) and bidimensional empirical mode decomposition (BEMD). The SWT could effectively reduce speckle noise; after an image was decomposed by BEMD, some bidimensional intrinsic mode function (BIMF) feature components were obtained, which could realize the expansion of gray difference between target and background. The SWT-BEMD algorithm not only improves target detection rate, especially those of small, hidden and weak scattering targets, but also reduces the influence of speckle noise and background clutter. The SAR image data verified the performance of the SWT-BEMD algorithm, and the experimental results show that it is effective and feasible.