SOFTWARE AND HARDWARE FOR DETERMINING GAUSSIAN NOISE LEVEL IN IMAGES
DOI:
https://doi.org/10.31891/CSIT-2022-1-6Keywords:
Gaussian noise, Gaussian filter, image filtering, convolution, Matlab, Simulink model, FPGA Artix-7Abstract
Accurate determination of the noise level in digital images is required to obtain their maximum signal-to-noise ratio, which is a necessary condition for the effective performance of the following stages of image processing: visualization, segmentation, recognition, etc. The task of calculating the Gaussian noise level is quite common, because such noise appears in most experimental images taken by video cameras. However, existing high-speed noise detection methods have a significant error, and the fairly accurate LLROI method has a low speed. The LLROI method is based on Low-frequency filtering of the noise component and Low-frequency filtering when selecting the Region Of Interest (ROI). Therefore, it is proposed to determine the level of Gaussian noise by the exact LLROI method and to increase its speed by appropriate hardware and software. Based on the LLROI method, a program in the MATLAB system was created, the structure and Simulink-model of a computer system for determining of Gaussian noise level on digital images were synthesized. Hardware implementation of image filtering units is made by FPGA Artix-7, which allowed us to increase the speed of the system. The results of calculating the Gaussian noise level for test images by the LLROI method using the developed hardware and software proved the errors not to exceed those provided by analogous methods.
The scientific novelty of the paper is to improve the LLROI method, namely to refine the threshold coefficient, which reduces the errors of calculating the noise level, even for images with clear contours and pronounced textures.
The practical significance of the developed tools is that they can be used to build high-speed computer systems (or subsystems) designed to increase the signal-to-noise ratio on digital images.
