Resolution enhancement of ultrasonic B-mode images

Abstract

IntroductionThe signal received from an ultrasound transducer contains information about the component on which the transducer is placed. When the amplitude of this signal is plotted on a colour scale against transducer position, a cross-section representation of the component is created; these are called B-scan images. B-scan images are an extremely useful tool to the NDT technician, allowing them to 'see' inside a component. B-scan images can be created by measuring the distance a single transducer has moved, or by using multiple element linear array transducers. As with many imaging systems, ultrasonic systems are not without their problems. As a consequence of the way sound waves propagate and behave, the resolution of the system will always be suboptimal. The oscillations of the crystal in the transducer create a wavelet, which contaminates the echo signals from the material under test. Also, due to beam divergence, echoes will be received from targets within the material under test that are not directly in the line of sight of the transducer. Due to this, the images created tend to exaggerate feature sizes and appear blurred.This paper proposes a three-step scheme of signal processing measures to increase the resolution of these images. The first step uses a deconvolution method to improve the temporal or through-thickness resolution of the acquired images. The proposed deconvolution method uses a threshold to determine safe areas within the spectra of the signals to ensure stability. The second step uses a new BMW-SAFT to improve the spatial resolution in the scan axis of the images. The BMW-SAFT algorithm uses the Fraunhofer approximation to model the far field of the ultrasound beam, which is then used to determine the size of the aperture and the weights applied across the aperture during the focusing process. The final step utilises a statistical method to remove background speckle noise within the images. Speckle noise can come from many different sources, but it is noted that after the first step of the proposed process, the deconvolution method creates speckle noise within the image as a side effect. To test the methods proposed in this paper, a validated computer simulation model has been developed to generate single A-scans and entire B-scans. The simulation model will be described in Section 2. Section 3 describes the three different processing methods proposed by this paper and compares them to their more conventional counterparts. In Section 4, a discussion on the performance and suitability of these methods is provided. Simulation model Forward modelAn ultrasonic inspection system can be considered to be linear time invariant. The transfer function for the material under test can be considered as a series of Dirac delta functions. Each Dirac delta function will represent a feature within the part under test. These features will include any interfaces where there is a change in acoustic impedance, such as the far surface of the part, any defect indications or reflections caused by the grain st...

Ultrasonic Defect Characterization Using the Scattering Matrix: A Performance Comparison Study of Bayesian Inversion and Machine Learning Schemas

“…While obtaining the crack propagation path and orientation by adopting TFM, Camacho et al showed that sizing by tip detection with phase coherence imaging (PCI) was more robust than the 6 dB drop approach [12]. In addition, signal and image processing techniques were developed which improve the time resolution of the measured A-scan signals [13], [14] and spatial resolution of the ultrasonic images [15], [16]. However, when the defect size is comparable to the ultrasonic wavelength, the defect indication becomes indistinguishable from that of a point scatterer and image-based characterisation of small defects is fundamentally challenging.…”