Tutorial 32 - Image filtering in python - Gaussian denoising for noise reduction

[@hacninetushar]

Your tutorials are very well planned and well structured. And also your teaching style is amazing. Thank you.

[@ZEISS_arivis]

You're very welcome!

[@DadaMichael-lq8qn]

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?

[@phaneendragarreprany4281]

Really wounderfull explanation

[@ZEISS_arivis]

Glad you think so!

[@oleksandraostapenko9076]

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.

[@ZEISS_arivis]

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.

[@oleksandraostapenko9076]

@@ZEISS_arivis Thank you for your answer!

[@elmirashyn6227]

Great videos and nice explanations.

[@ZEISS_arivis]

Glad you like them!

[@mariatsixla1450]

Thank you very much for your time and this video!

[@ZEISS_arivis]

My pleasure!

[@pinkhair4724]

Really good tutorial, thanks :D

[@vivek-159-icd]

as always interesting and informative video.

[@johngreen6186]

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?

[@ZEISS_arivis]

You can use imageJ to add artificial noise.

[@nickandoliu]

It's great. Thank you very much.

[@ananyapamde4514]

Beautiful lectures!

[@ZEISS_arivis]

Glad you think so!

[@David-kt5pj]

It's helpful. Thank you very much!

[@ZEISS_arivis]

You're welcome!

[@johngreen6186]

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

[@ZEISS_arivis]

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.

[@RaviYadav93]

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?

[@ZEISS_arivis]

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

[@PiyushSharma-bu3rq]

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.

[@ritujangra720]

sir how to implement gaussian filtering on the whole dataset of images....???

[@ZEISS_arivis]

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.

[@bandisrujanabhishek6122]

How to know what kind of noise is added in the image

[@ZEISS_arivis]

You don't know what type of noise is added, you make assumptions based on the source of the noise.

[@ritujangra720]

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

[@ZEISS_arivis]

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.