The meaning of NLD filtering is to compute the input image with a gaussian kernel. Its length is given by the sigma parameter.
The algorithm processes with a large gaussian kernel when we consider the background of the image, but it inhibits the filtering when it is about the edges, in order to keep important details of the image.
Lambda is a contrast parameter that depends on the image contrast. The best way to find a good value for it is to try different ones!
EED filtering is a different algorithm that filters the image by taking the edges orientation into account and thus keeps the edges more linear.
The image is, however, globally less filtered than in the NLD case.