Thresholding is transforming a gray-scale image to a black and white image
Disadvanteges of a normal averaging filter
Using an normal averaging filter could lead to losing some information because it will average from the real image and also the noise
Order-statistical filters
We have three types of order-statistical filter:
1. Median filter
Median filter is applied by taking the median of the values in filter dimensions, for example, let’s say we have an 3x3 area in the image with the following grey-scale values and a 3x3 median filter
1 & 2 & 6 \\ 12 &1 & 21 \\ 8 & 2 & 1 \end{bmatrix}$$ We first sort the values $$1, 1, 1, 2, 2, 6, 8, 12, 21$$ and then we choose the the median which is equal to 2 Median filters are really good with noise and pepper noise because the don't get averaged ,instead completely eliminated ### 2. Maximum filter ### 3. Minimum filter We don't take higher order sharpening filters because they will emphasize un-necessary changes that happen in the picture or image sum in sharpening filter must be 0 Sharpened images are given by doing a Laplacian filter and summing it to the original image only linear filters can be changed their order of application High boos filter gives more details to the center part Transfer function takes grayscale image to black and white and m is an hyper paramter where the transfer is shifted ![[Pasted image 20241009072257.png]] ## Gamma Correction ![[Pasted image 20241009072405.png]] gamma is an hyper parameter ![[Pasted image 20241009075941.png]] ![[Pasted image 20241009080311.png]]L doing integration to the histogram could be considered as a transfer function that would do gamma correction - maybe, this is just my interpretation ## Edge detectors They are filters that either use the first or second derivative, the sum of the filter must equal 0, so that when applied to a smooth picture the output is 0. ## TODO: fix this one