Unlike global thresholding, where a single threshold value is used to divide an image into two regions based on pixel intensity, adaptive thresholding uses a local threshold value that is computed for each pixel in the image based on the characteristics of its neighboring pixels.
This allows adaptive thresholding to handle images with varying lighting conditions, as the local threshold is adjusted to account for these variations.
There are several methods for adaptive thresholding, including Mean thresholding and Gaussian thresholding. Each method uses a different approach to calculate the local threshold value and has its own strengths and weaknesses.
Two options for thresholding are available, mean or gaussian.
The choice between Gaussian and mean thresholding for adaptive thresholding depends on the characteristics of the image being processed and the specific requirements of the application.
In general, Gaussian thresholding is more effective in dealing with images that have uneven illumination, as it uses a weighted average of the pixel intensities in the neighborhood of each pixel. This means that Gaussian thresholding gives more weight to the pixels closer to the center of the neighborhood and less weight to the pixels further away, which helps to account for any illumination variations.
On the other hand, mean thresholding is simpler to implement and computationally less expensive than Gaussian thresholding. It is therefore a good choice when processing large images or in situations where real-time processing is required.
Threshold
Defines the threshold mode, Binary or BinaryInv threshold value.
Kernel size
The kernel defines the region size around each pixel for calculating the threshold. As a rule of thumb, this should be larger than the objects to be counted. The size should not be set unnecessarily big, as this impacts the processing time very much.
NOTE: The adaptive threshold is most useful for small objects only. Using it on big objects makes it slow.