Lecture 4 pt 3

Question 1

What is a Bilateral Filter?

Answer 1

A Bilateral filter is a non-linear, noise-reducing, edge preserving filter.
It replaces the intensity of each pixel with a weighted average of intensity values from nearby pixels

Question 2

When are pixels considered close to each other in Bilateral filtering?

Answer 2

Pixels are considered to be close to each other if they occupy nearby spatial locations and if they have some similarity in the photo metric range.

Question 3

What parameters control bilateral filters?

Answer 3

There are two parameters to consider:
1. Range Parameter –> As it increases the bilateral filter gradually approximates Gaussian Convolution more closely.
2. Spatial Parameter –> As it increases it smooths larger features.

Question 4

Which is generally a better approach to use: Bilateral filter or Gaussian Convolution?

Answer 4

Bilateral filter as it’s range and spatial parameters provide more versatile control than Gaussian Convolution.

Question 5

What is the difference between Bilateral Filter and Gaussian Convolution?

Answer 5

Gaussian Convolution –> a local and linear filter that smooths the whole image irrespective of the edges or details.
Bilateral Filter –> A local non-linear filter that considers both graylevel similarities as well as geometric closeness of the neighbouring pixels without smoothing edges.

Question 6

What are the main applications of Bilateral Filters?

Answer 6

1. Denoising
2. Texture and illumination separation, Tone Mapping and Tone Management
3. 3D Fairing

Question 6

What are the main applications of Bilateral Filters?

Answer 6

1. Denoising
2. Texture and illumination separation, Tone Mapping and Tone Management
3. 3D Fairing

Question 7

How is Bilateral filtering implemented?

Answer 7

Bilateral Filtering is implemented in the frequency domain by multiplication.
The image in the frequency domain is multiplied by the filter’s frequency response in the frequency domain.

Question 8

What are the main advantages of Frequency Domain Filtering?

Answer 8

1. the process is simplified to a multiplication, rather than convolution.
2. It gives you more control over the whole image.
3. Tools used for the image processing like correlation, convolution etc are much simpler and computationally cheaper.

Question 9

What is a period or wavelength?

Answer 9

Units necessary to describe one cycle in the case of images that are space units. e.g. pixels