This paper presents a wavelet-based image processing technique, which analyzes eddy-current testing (ECT) images derived by scanning printed circuit boards (PCB's) with an ECT probe and automatically detects the existence and location of the defect. First, the undesired components contained in probe output are removed through two types of wavelet filtering. Then the comparison of two images obtained from reference and tested PCB's are carried out to extract the signal due to the defect. In this paper, one-dimensional (1-D) wavelet is used only in the horizontal direction considering that the scanning of the probe is along that direction. In addition, the square norm of difference between original and processed signal is proposed as a criterion to keep the waveform of the defect peak as possible. The application examples of sample PCB's reveal the effectiveness and problems of the given approach.