All naturally occurring crack-like defects in solid structures are rough to some degree. Based on simulated array data for incident and scattered longitudinal waves from various rough cracks and the total focusing method (TFM) imaging algorithm, the effect of roughness on defect imaging and characterization is examined. The array data are simulated by using a previously validated model based on the Kirchhoff approximation to predict the rough crack scattering matrix. The scattering matrix describes the scattered field for all possible incident and scattering directions. The performance of TFM imaging for detecting rough crack-like defects is investigated by considering the averaged scattering coefficient over specified angular coverage ranges and the maximum scattered amplitude in the final image. It is shown that roughness can be either beneficial or detrimental to the detectability of a crack-like defect, depending on the defect characteristics, such as length, roughness, correlation length, orientation angle, and array inspection configuration. It is also shown that roughness can cause underestimation of the crack length if using an image-based approach for sizing.