Spatial Kernel size - Gabor filter bank

Hi

Does anyone know how the kernel size in the Gabor function is calculated from input data such as wavelength and direction?

https://la.mathworks.com/help/images/ref/gabor.html#buuu1h2-1-wavelength

Answer 1

[Balavignesh]

Hi Paola,

As per my understanding, you would like to find out how 'kernel size' is calculated from the input data in the 'Gabor' function.

I would suggest you choose an appropriate aspect ratio ('AR') for the kernel size and calculate the spatial frequency ('f') from the wavelength.

You could then determine the standard deviation of the Gaussian envelope ('p') from the desired spatial frequency and aspect ratio. The kernel size could be calculated from the standard deviation using the following relation:

'kSize = ceil(3 * p) * 2 + 1'

The following code may help you achieve this:

% Sample input values. Use your desired values 
wavelength = 4;   
orientaion = 0;   

% Calculate spatial frequency 
frequency = 1 / wavelength; 

% Set aspect ratio 
aspectRatio = 0.5; 

% Calculate standard deviation 
sigma = 0.56 * (wavelength * aspectRatio + 0.25) / (wavelength * aspectRatio - 0.35); 

% Calculate kernel size 
kernelSize = ceil(3 * sigma) * 2 + 1 

Please refer to the following documentation links to have more information on:

• 'ceil' function: https://www.mathworks.com/help/matlab/ref/ceil.html?searchHighlight=ceil&s_tid=srchtitle_support_results_1_ceil
• 'gabor' function: https://in.mathworks.com/help/images/ref/gabor.html?searchHighlight=gabor&s_tid=srchtitle_support_results_1_gabor

Hope that helps!

Balavignesh