Lecture 4 pt 2 Flashcards
What are local (neighbourhood) operations?
Image operations that consider the neighbouring pixels of a coordinate. (The rectangle around a central pixlel known as the Kernal or Filter
What is a linear filter
A linear filter is the most common neighbourhood filter.
It’s output pixel’s value is determined as a weighted sum of the input pixel values. It also uses a convolution operator
What is padding?
Deals with boundary effects. There are 3 types of padding:
1. Zero
2. Constant
3. Clamp
4. (Cyclic) wrap
5. Mirror
6. Extend
What is Zero Padding?
Zero Padding is when all pixels outside of the source image are set to 0
What is Constant padding?
Sets all pixels outside of the source image to a specified border value
What is clamp padding?
Clamp padding is having the edge pixels repeat indefinitely.
What is cyclic / wrap padding?
Loop around the image in a toroidal configuration.
What is Mirror padding?
Mirror padding reflects pixels across the image edge
What is extend padding?
Extend the signal by subtracting the mirrored version of the signal from the edge pixel value
What is Box Filtering?
Box filtering is a type of linear filtering. It averages the pixel values in a k x k window. Each pixel is computed as the average of the surrounding pixels.
What is Gaussian Filtering?
Gaussian Filtering is a type of linear filtering.
It takes a square array of pixels where the pixel values correspond to the values of the Gaussian Curve (Bell Curve)
What are Low-Pass Filters
Low-Pass filters are employed to remove high spatial frequency noise from a digital image.
They usually employ moving window operators which affects one pixel of the image at a time, changing its value by some function of a local region (window) of pixels.
What are non-linear filters?
Similar to linear filters, but instead of using the spatial mask in the convolution process, the mask is used to obtain the neighbouring pixel values, and then ordering mechanisms produce the output pixel.
