In this paper, an extensive research on grayscale images applying spatial domain filtering technique to remove additive white Gaussian noise (AWGN) has been developed. However, the existing average based filters have a main limitation which is the exploitation of the statistics of the original noisy image only. Nevertheless, in the infinite number of looks prediction (INLP) technique, in addition to the statistics of the original image, the statistics of the filtered images are also used to elevate the filtering performance. A linear regression between the filtered pixels and their variances for various window sizes has been achieved. The aim of this work is to extent the INLP filter, which is originally used with multiplicative noise, to AWGN image filtering. A theoretical effort has been conducted to derive a linear rule between the filtered pixels and their variances of AWGN digital images. Then, a linear regression has been applied to compute the INLP filtering pixels. Experiments carried out on gray-level images show the potentiality of the approach compared to other established algorithms, in terms both of noise reduction and spatial detail preservation.