Non-destructive testing methods are used largely in component manufacturing industries like Aerospace, Renewables and Power to evaluate the properties of a material or the quality of a component by inspecting for cracks and discontinuities without causing damage to the part. Among the many non-destructive testing methods, Eddy current imaging enables efficient flaw detection for surface and sub-surface cracks. However, in typical eddy current inspection, there can be significant number of false calls arising from variation in lift-off and surface anomalies. Discriminating defect signals from false calls can be very challenging. This paper describes a method to reduce false calls by using a wavelet based denoising algorithm and combining it with statistical-based features extracted inside a sliding window in the time domain to efficiently identify the cracks. The results are verified on specimens with cracks of different sizes that are oriented randomly along with locations available for baseline noise measurements.