In the last few decades huge amounts and diversified work has been witnessed in the domain of de-noising of binary images through the evolution of the classical techniques. These principally include analytical techniques and approaches. Although the scheme was working well, the principal drawback of these classical and analytical techniques are that the information regarding the noise characteristics is essential beforehand. In addition to that, time complexity of analytical works amounts to beyond practical applicability. Consequently, most of the recent works are based on heuristic-based techniques conceding to approximate solutions rather than the best ones. In this chapter, the authors propose a solution using an iterative neural network that applies iterative spatial filtering technology with critically varied size of the computation window. With critical variation of the window size, the authors are able to show noted acceleration in the filtering approach (i.e., obtaining better quality filtration with lesser number of iterations).