One or more systems and/or techniques are provided to identify and/or classify objects of interest (e.g., potential granular objects) from a radiographic examination of the object. Image data of the object is transformed using a spectral transformation, such as a Fourier transformation, to generate image data in a spectral domain. Using the image data in the spectral domain, one or more one-dimensional spectral signatures can be generated and features of the signatures can be extracted and compared to features of one or more known objects. If one or more features of the signatures correspond (e.g., within a predetermined tolerance) to the features of a known object to which the feature(s) is compared, the object of interest may be identified and/or classified based upon the correspondence.