Convolution Filter

This class modifies the content of the GLI frames. All the frames that "pass" through this class will be filtered using a convolution filter.

Add the "Filter Conv" class in the "Class management (2D)"

mode and press "Update DB" to activate the new class.

A new "Filter Conv" mode will be available from the "Mode" menu

The available filters are:

		No Filter The image is not affected by the class

		Median For each pixel i,j in the original image, the resulting pixel will be computed by taking the median value of the pixel and its neighbors.

		Mean For each pixel i,j in the original image, the resulting pixel will be computed by taking the mean value of the pixel and its neighbors

		Gaussian The resulting image is produced by convolving the source image with a Gaussian kernel:

		Laplacian of a Gaussian (LoG) The resulting image is produced by convolving the source image with a kernel formed by the Laplacian of a Gaussian :

		Sharpen The resulting image is produced by convolving the source image with the kernel of profile:

		Laplace The resulting image is produced by convolving the source image with the kernel:

		Roberts The resulting image is produced by convolving the source image with the two kernels A and B and adding the results: res = ABS[A] + ABS[B]

		Sobel The resulting image is produced by convolving the source image with the two kernels A and B and adding the results: res = ABS[A] + ABS[B]

The Median, Mean and Gaussian filters are used to reduce the noise in the images. The Mean and Gaussian filters will blur the edges while the Median filter will maintain clean edges.

The Sharpen and Laplace filters are contrast enhancement filters. These are useful when you would prefer to see anatomical structures more clearly.

The Roberts and Sobel filters are edge detection filters designed to increase the visibility of the edges in the images.

From the Graphic Interface


Instance selection	Select which of the instances of the class you want to work on.
Filter list	Clicking on one of the filter buttons will select the corresponding filter. The available filters are: Median, Mean, Gaussian, LoG, Sharpen, Laplace, Roberts and Sobel. A brief description of each filter, with some image examples, is located above.
Kernel Size	You can adjust the Kernel Size of the Mean, Median, Gaussian and LoG filters.
Demo images	Three demo images are provided to help in the filter and mix selection. The image on the left is unfiltered, the image on the right is fully filtered and the center image is a mix of these. The mix parameter is controlled by the mix slider.
Mix slider	Filtering operations are controlled by the "mix" slider. The filtered image is mixed with the original image in a proportion that is controlled by the mix slider. If the slider is completely to the left, the filtering operation will not affect the image. If the slider is completely to the right, the image will be replaced by its filtered version.
Update DB	Clicking on this button will apply this filter to the selected images.

From the Display Area

There is no display area interaction specific to this class.

From the Keyboard

There is no keyboard interface specific to this class.

From the Command Line

There is no command line for this class.

Technical Note

Sharpen, Laplace, Roberts and Sobel filters are very sensitive to the noise in the image. You may want to precede these operations with a noise reducing filter (Median, Mean, Gaussian).