Efficient segmentation of optoacoustic images has importance in enhancing the diagnostic and quantification capacity of this modality. It may also aid in improving the tomographic reconstruction accuracy by accounting for heterogeneous optical and acoustic tissue properties. In particular, when imaging through complex biological tissues, the real acoustic properties often deviate considerably from the idealized assumptions of homogenous conditions, which may lead to significant image artifacts if not properly accounted for. Although several methods have been proposed aiming at estimating and accounting for the complex acoustic properties in the image domain, accurate delineation of structures is often hindered by low contrast of the images and other artifacts produced due to incomplete tomographic coverage and heuristic assignment of the tissue properties during the reconstruction process. In this letter, we propose instead a signal domain analysis approach that retrieves acoustic properties of the object to be reconstructed from characteristic features of the detected optoacoustic signals prior to image reconstruction. Performance of the proposed method is first tested in simulation and experiment using two-dimensional tissue-mimicking phantoms. Significant improvements in the segmentation abilities and overall reconstructed image quality are further showcased in experimental cross-sectional data acquired from a human finger.