Getting rid of noise in an image

I am working in segmentation, edge detection and opening and closing images by reconstruction with the aim of extracting circular objects of a certain radius from a noisy image. Doing opening and closing by reconstruction and generating a binary mask along with using the watershed function allowed me to somewhat do segmentation and obtain some of the circular objects that I need to extract. However, using the edge detection function canny gives me better results in terms of identifying the edge of the circular objects I want. The problem is that the canny function also draws the edges of a lot of the noise in my image and that prevents me from extracting the circles properly. The final idea is to recognize the circles very well in the image in order to create a mask with the circle's edge that I can then superimpose in the original noisy image. The question is can anyone suggest a way of getting rid of all the noise picked up by the canny function and get a mask with only the edge of the circular object? or a way of extracting circles of certain radius from a noisy image? P.S. imfindcircles does not work because my image is quite noisy. Attached is the original image. I appreciate any help