Digital images are often contaminated by noise during image acquisition and transmission. Noise degrades the image quality and causes some loss of information details. It is very important to suppress these noises in images before subsequent processing, such as image segmentation, edge detection, etc. But in practical applications, noise in the images is often a mixed one, which might be composed of Gaussian noise and impulse noise. Unfortunately, the existing denoising algorithms are often designed to deal with a single kind of noise, Gaussian noise or impulse noise. For the mixed noise, those algorithms can not achieve satisfactory effect. In this paper, a new method which can efficiently suppress the Gaussian-impulse mixed noise is proposed. This method is described as follows. Firstly, the boundary discriminative noise detection (BDND) algorithm is used to detect the pixels corrupted by the impulse noise and a two-dimensional binary decision map is formed after the detection stage. Secondly, the noisy image is filtered according to the decision map by the bilinear interpolation filtering algorithm. As a result, those pixels corrupted by the impulse noise are corrected and a temporary image corrupted by the Gaussian noise is obtained. Finally, the existing Bayes least squares-Gaussian scale mixture (BLS-GSM) algorithm is used to denoise the temporary image. Experimental results show that the proposed method can efficiently remove the Gaussian-impulse mixed noise. |