CPC G01V 1/364 (2013.01) [G06T 5/20 (2013.01); G06T 5/50 (2013.01); G06T 5/70 (2024.01); G01V 1/34 (2013.01); G01V 2210/324 (2013.01); G06T 2207/20081 (2013.01); G06T 2207/20084 (2013.01)] | 18 Claims |
1. A method for processing an image, comprising:
receiving an input image of actual data comprising a signal and noise, wherein the actual data is recorded by one or more sensors;
generating a filtered image based on the input image of the actual data by removing at least a portion of the noise from the input image, wherein a portion of the signal is also removed from the input image;
generating a residual image based on the input image of the actual data, wherein the residual image comprises the at least a portion of the noise and the portion of the signal that are removed from the input image to generate the filtered image;
identifying at least some of the portion of the signal of the actual data that is in the residual image at a pixel-by-pixel or voxel-by-voxel level, wherein the at least some of the portion of the signal is identified using a machine-learning model trained by a plurality of images comprising one or more coherent shapes representing the signal to distinguish between signal and noise in the residual image;
inserting the at least some of the portion of the signal of the actual data identified in the residual image into the filtered image as pixels or voxels;
detecting one or more discontinuities in the filtered image that are or would be caused by inserting the at least some of the portion of the signal into the filtered image, wherein the detecting occurs both before and after the inserting; and
adjusting the filtered image or the at least some of the portion of the signal to mitigate the one or more discontinuities.
|