Cross Correlation Flashcards
What is the process of using a filter kernel on a selection of pixels?
- Select a filter kernel
- For each pixel in the image
- Can the kernel be centred over the pixel
- Calculate the sum of products and replace the current pixel with the new sum
- Once all pixels are done normalise the new pixels
What do filter kernels create as a side effect?
Edges, where filter could not be centred
What are some counters to the edge problem?
- Reduce size of image
- Put a single colour border or pad around the image beforehand
- Use old pixel values
- Use known neighbours colour
What is the intermediate value equation?
I’ = ((I-min)*255) / (max-min)
What is normalisation?
The process of calculating the intermediate values
How does the filter kernel work for coloured images?
Calculate the sum of products using the kernel filter for each colour independently, then normalise with max and min overall value.
What is correlation?
A form of spatial linear filtering
How can we spot an edge on graphs?
Where that “rate of change”/gradient intensity is large
What are derivatives used for?
To find where the rate of change of intensity is highest
Where are edges found on a 2nd derivate graph?
When the function crosses from positive to negative.
What is edge detection using the Sobel operator?
Uses a x and y filter and then calculates the magnitude and direction of the gradient. If the gradient is greater than some grey value map it to 0-1.
How does median filtering work?
Works by replacing each pixel value in an image with the median of the pixel values in a given neighbourhood. Size of neighbourhood determined by kernel size
What is the advantage of median filtering?
It’s very effective at removing high/low intensity noise and random noise
What is the disadvantage of median filtering?
It can cause blurring and loss of fine detail if the kernel size is too large
How does Gaussian smoothing work?
Uses a Gaussian kernel on an image, which assigns more weight to nearby pixels than to distant pixels.
What is the advantage of Gaussian smoothing?
Effective in removing gaussian noise or random noise from an image. Smooths noise without blurring edges or detail
What is the disadvantage of Gaussian smoothing?
It can blur the edges and detail if the kernel size is too large or noise is too high
