The impact of Gaussian filtering on image clustering accuracy

The influence of the results of image processing by a Gaussian filter with a matrix of 11 by 11 is considered. For this purpose, a dataset of 77 color images containing photographs of road, cars, buildings and town is selected. The images are converted to shades of gray, and then a Gaussian filter is applied to the images. The resulting and original images are clustered using the k-means and c-means algorithms. It is concluded that both methods give approximately the same values of the accuracy of clustering the blurred images by the IoU metric. It is found that the Gaussian filter changes the clustering of images by about 10%.