Tutorial 32 - Image filtering in python - Gaussian denoising for noise reduction
Вставка
- Опубліковано 24 сер 2024
- In microscopy, Gaussian noise arises from many sources including electronic components such as detectors and sensors. In addition, salt & pepper noise may also show up due to errors in analog to digital conversion. Therefore, image denoising is one of the primary pre-processing operations that a researcher performs before proceeding with extracting information out of these images.
This tutorial explains Gaussian denoising filter and walks you through the process of writing a couple of lines of code in Python to implement the filter.
Code associated with these tutorials can be downloaded from here: github.com/bns...
Your tutorials are very well planned and well structured. And also your teaching style is amazing. Thank you.
You're very welcome!
Really wounderfull explanation
Glad you think so!
Great videos and nice explanations.
Glad you like them!
All your tutorials are amazing and easy to follow. Many thanks for making them available for all. Have you attempted any form of model deployment with Gaussian denoising?
Really good tutorial, thanks :D
as always interesting and informative video.
Thank you very much for your time and this video!
My pleasure!
Beautiful lectures!
Glad you think so!
It's great. Thank you very much.
It's helpful. Thank you very much!
You're welcome!
Sir you have added 25 sigma gaussian noise in first image and secondly how much noise is added in salt and pepper noise image.
If we are calculating the efficiency of different filters then equal amount of noise should be added in gaussian, poisson, salt&pepper and speckle image so that they can be compared on same scale.
Thank you for the tutorial!
Could you please explain why you chose for the Gaussian smoothing example (5:06) the kernel which is not Gaussian? The kernel values do not go down symmetrically.
The first part of the presentation explains Gaussian. At around 2.53 you can see the Gaussian kernel on the screen. The part around 5.06 is about convolution and what it means by using random kernel, not Gaussian. I see where it may have confused you, should have used a Gaussian kernel.
@@ZEISS_arivis Thank you for your answer!
In 6:48, the first two images you're reading there are already images that you added to them gaussian noise and salt and pepper noise?
where can i get these images or can you say how you apply these two operations on a given image?
You can use imageJ to add artificial noise.
You have defined the size of a kernel in Opencv but there is no argument in skimage to change the kernel size. Then, how to change the size of the kernel in skimage?
skimage uses sigma, the standard deviation of the gaussian kernel instead of actual kernel size. scikit-image.org/docs/dev/api/skimage.filters.html#skimage.filters.gaussian
1) what's the goal of the filter you're using in 4:30 + what's the meaning of replacing pixel value 42 with 394..what's the goal/effect of this
convolutional operation here?
2) regarding the kernel/filter you're using, it's called gaussian kerenl but why? how it relates to the gaussian formula in 1:42?
thanks and btw, ur videos are great
These are questions about digital images and it is necessary background to be able to work with images. When we replace a pixel value we are literally replacing an existing value with a new value. This is basically image processing, when you process an image you are working with pixel values. By performing convolutional operations you are doing math at every pixel. For Gaussian kernel, the matrix is defined to represent Gaussian shape.
sir cud u pls make a video elaborating reading of image dataset and applying filtering on it and then applying some feature extarction on the same datset and then feature selection.......cud make one video trying all these things on one image datset....it wud b a great help
Please keep watching this channel you may find examples explaining what you are looking for. The tutorials about image segmentation will definitely cover part of it.
How to know what kind of noise is added in the image
You don't know what type of noise is added, you make assumptions based on the source of the noise.
sir how to implement gaussian filtering on the whole dataset of images....???
If your images are present in a local directory you can use glob or os.listdir to walk through each image and process it. Please watch videos 27 and 28.