A spatial filtering algorithm in low frequency wavelet domain for X-ray inspection image de-noising

Image de-noising algorithm is the crux of digital Xray inspection for inner defects in electronically package production. Due to various factors the actual noise derived from the real X-ray inspection images in site have a certain space-time correlation or peculiar characteristics of colored noise, specifically, the frequency distribution of the noise in the frequency domain by Fast Fourier Transform(FFT) is not average but concentrated within a relatively low frequency band. Based on the low frequency characteristics of the noise, the spatial filtering the wavelet transformed X-ray images in the low-frequency region is desirable. The integrated spatial filtering algorithm introduced in this paper includes three sections as follows: Firstly, a layer of wavelet transform is applied to the original X-ray inspection image; secondly, an improved mean filtering algorithm based on gray-scale difference is exerted to the upper-left image which is in low-frequency domain of the wavelet transformed image; thirdly, using wavelet inverse transform algorithm to restore the X-ray inspection image. Comparing the two amplitude-frequency diagrams by FFT of the noise being respectively from the before and after spatial filtered images, the amplitude in the medial region of amplitude-frequency diagram by FFT of the latter noise is decreased obviously. This states that the spatial filtering algorithm of low frequency wavelet domain can be effective to the noise reduction and image quality improvement for the X-ray inspection image.