Linear Filters Flashcards
How does the process of spatial filtering differ from point processes?
Change is made to one pixel at a time BUT a whole window/kernel surrounding the source pixel is considered
What is image noise?
Small errors in image values
How is noise often modelled?
recorded value = true value + random noise value
What are the two main ways noise is added to an image?
During image capture, and during image compression
What is the concept of gaussian noise?
Sensors often measure a value as a little off, but on average they are right, and generally the values are closer to true rather than further. So we can model noise as a gaussian distribution
What is the mean added noise value?
0
What is convolution filtering?
A process where a filter is run over parts of an image covering a window at a time
What are some negatives of using the mean filter for noise reduction?
Hard to use around the edges where there are not as many surrounding pixels
What is gaussian filtering?
Convolution filtering but with a mask whose weights are determined by a 2D gaussian function (e.g higher weight is given to pixels closer to the source pixel)
What does the volume under a 2D gaussian add up to?
1
What does a gaussian filter look like?
A grid of numbers with the largest in the middle
