Ultrasonic techniques are able to accurately detect and characterize flaws in homogeneous structures. Elastic reverse time migration (ERTM) is a powerful tool to reconstruct high-resolution images of flaws. To achieve images with better quality, the solution can be obtained by iteratively finding an image generating the modeled data which can best match the measured data in a least-squares sense, i.e. least-squares migration (LSM). Combing ERTM and LSM, conventional elastic least-squares reverse time migration (ELSRTM) methods are based on the assumption of a constant density, which can lead to inaccurate amplitudes and parameter crosstalk artifacts in the reconstructed images. In this paper, an ultrasonic imaging technique based on the ELSRTM which considers density as well as longitudinal-(L-) and shear-wave (S-wave) velocity variations is explored for imaging flaws in heterogeneous structures. The ELSRTM with density variations can simultaneously reconstruct density and L- and S-wave velocity images, which can provide amplitude-preserving images and mitigate crosstalk artifacts. This method is applied to numerical as well as physical laboratory experiments and the results appear promising for flaw identification in heterogeneous structures