It is interesting to compare the output of the 7x7 box blur shown on this slide, with the output of the 7x7 truncated gaussian blur filter shown on slide 22. What do you notice?
Advanced question: If know a little bit about the properties of the box filter and the gaussian filter in frequency domain (we didn't talk much about these properties in class), how might you describe the differences in the output images in terms of those properties?
Haloee
It's blurred but not very smooth. There are "jumps" between the grey edges and red block. Also there are horizontal white bars.
Gaussian filter preserves better edge information and let the overall blur be smooth. I think the reason is that when applying Gaussian filter, by weighting center of the kernel more, the edge pixel will reserve more information provided by itself rather than be over interfered by the pixels around it.
It is interesting to compare the output of the 7x7 box blur shown on this slide, with the output of the 7x7 truncated gaussian blur filter shown on slide 22. What do you notice?
Advanced question: If know a little bit about the properties of the box filter and the gaussian filter in frequency domain (we didn't talk much about these properties in class), how might you describe the differences in the output images in terms of those properties?
It's blurred but not very smooth. There are "jumps" between the grey edges and red block. Also there are horizontal white bars.
Gaussian filter preserves better edge information and let the overall blur be smooth. I think the reason is that when applying Gaussian filter, by weighting center of the kernel more, the edge pixel will reserve more information provided by itself rather than be over interfered by the pixels around it.