On processing noisy contrast images

The problem of noise reduction at sharp transitions of brightness in digital noisy contrast images is considered. In addition to the useful signal, digital images obtained by digitizing an analogue signal with a digital photo matrix have a noise component. Moreover, to obtain a digital image in the standard RGB color model, a demosaicing interpolation algorithm must be applied to the image obtained from a digital photo matrix. Due to such transformations, the Gaussian distribution of noise in a digital noisy image is violated. Using a standard image digitization model for noise reduction is not effective. For more effective noise reduction, the digital image is transferred from the RGB color model to the HSV or LAB color model, where the brightness and color components of the digital noise can be filtered separately. Color noise is suppressed in the color channels of the image using a Gaussian filter. Noise reduction in the brightness channel of a digital image is more difficult task, especially at the edge of sharp transitions of brightness. To suppress the brightness noise in contrast images, it is proposed to use a nonlinear filter based on the generalized method of least absolute values (GMLAV). The process of smoothing the contrast noisy transition by the GMLAV-filter is described, and its efficiency is shown in comparison with the median filtration.