Ok

You are welcome!

I will record a video on how to work with 3D images. Until then, here is some info... 2D images have numpy array shape as (x, y) for gray images and (x, y, RGB) for color images. 3D images typically have structure (z, x, y) for gray and (z, x, y, RGB) for color image stacks. Most opencv and scikit-image functions are designed for 2D images. Therefore you need to read one slice at a time inside a for loop and apply the image processing function. Easier than it sounds... If I have an image of size (50, 256, 256) - 50 slices gray. If I want to apply threshold to the entire volume, slice by slice. binary_img = [] for img in range(image.shape[0]: #This would be 50 in our example input_img = image[img, :, :] #Define temporary image slice name thresh = 75 #Or use otsu binary = input_img > thresh binary_img.append(binary) #Add binarized slice to our placeholder list. processed_img = np.array(binary_img) #Convert to numpy array #Now save to a tiff file if you want.

@@DigitalSreeni Thanks for the reply. My concern is that is processing images slice by slice may be not the same as using a real 3D filter. For example, Imagej provides both 2D and 3D Gaussian filters. In 3D Gaussian filters, sigma can be assigned to x, y and z. However, in Scikit-image, it seems that only one sigma can be defined. Also, I think 2D watershed can not assigned the same grain in different slices the same label as 3D watershed does. From Scikit-image docstring, its watershed works for both 2D and 3D images but not sure how it works.

If you want to clean dead pixels you can use median filter. If you want to detect the location of dead pixels then you need to write code to run median filter and identify hot spots. I do not have anything ready on this topic.

In almost all cases, you assume that noise is unstructured, which means it is random. This gives you a few choices. I recently gave a keynote presentation on this topic, here is the link. kzfaq.info/get/bejne/r7VhaKx6uL6lYY0.html

@@DigitalSreeni Thank you so much sir

Use glob or os.listdir to read multiple files from a directory and applying a function. I recorded a couple of videos for my work channel, may be they can help. kzfaq.info/get/bejne/kJ9gfqiir9WnY6s.html kzfaq.info/get/bejne/oJx3gdei193Hfnk.html

@@DigitalSreeni Sir it gives error when I apply filter to os.listdir(). So can you please give me code for that. Actually I am not able to apply the function to os.listdir

@@DigitalSreeni I have 1 folder which 8 folders in it and each one contains images. Actually I have a multiclass dataset. So I want to apply denoise function to training dataset as well as validation dataset. So please give a code for that.

Did you try medpy library, I remember them having a 2D version of that filter.