A novel modi ed level set method (MLSM) is proposed to reconstruct the shape and inhomogeneity of two dimensional objects. As a numerical technique, level set method is capable of revealing the shape and position of objects by synthetic/measurement data. In fact, the constitutive parameters of an object (e.g., its relative complex permittivity) are among the apriori information for this inversion algorithm. In the present method, an evolution strategy is proposed so that both shape and permittivity of a temporary object are evaluated. The temporary object converges to the target while a cost function is decreasing. The cost function is regularized with two penalty terms. In order to prevent a sudden change in the shape of the temporary object, a curvature-based regularization is used. Also, the Laplacian regularizer is adopted to reduce the uctuation and roughness of the constitutive parameters of the temporary object. Using di erent synthetic data, the capability of MLSM in accurately identifying objects of interest is presented. It is shown that with MLSM, we are able to completely separate the two adjacent objects with in a distance more than one fteenth of a wavelength. Key Words Microwave Imaging, Inverse Scattering, Level Set Method, Linear Sampling Method