METHOD OF SUPPRESSING NOISE ON X-RAY IMAGE USING ARTIFICIAL NEURAL NETWORK Russian patent published in 2024 - IPC G06T1/40 G06N3/08 

FIELD: data processing.

SUBSTANCE: invention can be used to suppress noise on an X-ray image. Summary of the invention consists in the fact that a series of initial X-ray images is obtained using an X-ray receiver and a noise generator (NG) is calibrated, training an artificial neural network (ANN) using a training data set consisting of a plurality of pairs of an X-ray image (XI), each of which includes a noisy XI and a reference XI, obtaining a noise-free XI using a pre-trained ANN and outputting the XI to a user or for further processing, wherein the training of the ANN is carried out by alternating the training stage using artificial noise and the training stage using real noise, each of which is carried out iteratively until the convergence condition is satisfied, wherein at the training stage using artificial noise at each iteration, XI is supplied to the ANN input with artificial noise added to them, calculations are performed inside the ANN structure, calculating the error between the ANN output and the corresponding reference XI and using the error back propagation method, updating the ANN weights, at the training stage using the real noise at each iteration, the initial XI are supplied to the ANN input, each of which is taken from one of the XI series, each of which is obtained by shooting a static object in constant exposure conditions, repeated N times, where N ≥ 2, calculations are performed inside the ANN structure, calculating an error between the output of the ANN and the corresponding reference XI and updating the weights of the INS using the error back propagation method.

EFFECT: providing the possibility of improving the quality of the obtained X-ray images.

8 cl, 7 dwg