Centrifugally cast stainless steel (CCSS) components are currently being used in many critical industrial applications such as nuclear reactors, which have stringent inspection requirements. Non-destructive testing (NDT) of these components are governed by strict guidelines to insure safe operating conditions. Current literature indicates that ultrasonic NDT techniques provide potentially the most promising and reliable methods for the inspection of CCSS components. However, the received signal often has low signal-to-noise ratio (SNR) due to ultrasonic attenuation caused by grain scattering. Furthermore, ultrasonic inspection of CCSS components is plagued by high attenuation, velocity variations, mode conversion, beam divergence and/or convergence, and skewing. Therefore, the signals arising from metallurgical discontinuities or defects caused by thermal fatigue and stress corrosion, can appear to a manual ultrasonic inspector as random, stationary-noise signals. The ability to detect such defects is at best limited with the conventional non-destructive evaluation techniques. In this paper, some spectral methods are presented to improve flaw visibility by reducing the background noise from the CCSS microstructure.