Method of noise reduction in digital X-rayograms
The invention relates to the field of digital image processing and can find use in suppression of noise in digital images, formed by high-energy radiation, including X-ray radiation. Specifically, the invention relates to a method for suppression of noise in digital x-ray images. The objective of the invention is to provide a method for noise suppression in digital X-ray images that has extended functionality, specifically, a method that makes it possible to reduce residual noise level and amount of artifacts in the form of discontinuities, directed along local orientation of object borders in textured image areas, to reduce residual LF noise level, and to eliminate over-smoothing (excessive filtering) of fine details. The technical innovation achieved is the improvement of digital image processing quality.
1. A method for noise suppression in digital X-ray images, comprising:
acquiring an original image;
estimation of a signal-dependent noise variance dependent on signal intensity;
combined bilateral non-local filtering;
suppressing residual low-frequency noise; and
over-smoothing correction,
wherein the estimation of a signal-dependent noise variance dependent on signal intensity comprises:
morphological elimination of the values of noise image pixels corresponding to edges in the original image;
using robust piecewise linear approximation of interval estimations of noise variance to obtain a tabulated function of dependence of noise on signal strength; and
calculating a noise map as a pixel-by-pixel estimation of noise variance in the original image;
wherein the combined bilateral non-local filtering comprises averaging by combining non-local means filter and bilateral filter weights;
wherein suppressing residual low-frequency noise comprises using a pyramid method with filtering of low-frequency by a bilateral filter;
further comprising obtaining a result image; and
wherein the over-smoothing correction comprises combining the final image and the original image based on structural and pixel similarity between the filtered image and the original image.
2. The method of claim 1 , further comprising calculating the averaging weights for combined bilateral non-local filtering using the original image smoothed by a linear low frequency filter.
3. The method of claim 1 , wherein the combined bilateral non-local filtering comprises using a combined weight function as fixed weight fractions for the non-local means filter and the bilateral filter, wherein these filters are combined at a final stage.
4. The method of claim 1 , wherein the combined bilateral non-local filtering comprises using a separable version of the bilateral filter for separation of the non-local means filter and the bilateral filter.