INVESTIGATION OF THE STABILITY OF RADON-TO-NOISE CONVERSION DURING LINE DETECTION IN IMAGES

This paper considers modifications of the Radon transform aimed at detecting straight lines in noisy images. A comparative analysis of three approaches is presented: the classical Radon transform, the thresholded modification, and the localized integration. The thresholded approach suppresses low-intensity values prior to transformation, while the localized method limits the integration region to a finite-length segment. Each method was tested on synthetic images with varying levels of noise. Based on the analysis of peaks in the sinogram and the reconstructed lines, it is shown that each modification has its own advantages and is applicable under different conditions. The study highlights the relevance of the problem and the need for further research on the Radon transform in the context of line detection in images.