Image Scaling with Edge Detection using Bilateral Filters

1
Image Scaling with Edge Detectionusing Bilateral
Filters
Haarith Devarajan, Harold Nyikal
2
Outline

• Motivation
• Image Scaling
• Edge Detection
• De-noising
• Bilateral filtering
• Process
• Results

3
Motivation

• Up-scaling of images is done by up-sampling
  pixels.
• The values of new pixels are obtained by
  interpolation of adjacent pixel values
• Low-pass filtering.
• Sharp edges are not preserved and image appears
  blurred.
• We need to up-scale while preserving edges

4
Image Scaling

• We focus on up-scaling
• Common technique
• Up-sample by adding zeros between pixels
• Interpolate between pixels
• Interpolation methods
• Nearest neighbor blocky edges
• Bilinear interpolation blurred edges
• Bicubic interpolation blurring edges
• Blurring proportionally increases with order of
  up-scaling

block edges
blurring
5
Edge Detection
Input
Scaled Output
On scaling
Detect edges
Better Response !
On scaling
6
Edge detection algorithm

• Second order derivative
• Good for edge detection
• But, also amplifies shot noise
• We need to limit noise before/after edge
  detection scaling

7
Denoising

• Use low pass filters
• Uniform filter
• Bartlett filter
• Gaussian filter
• Sinc filter

8
Bilateral Filtering
Edge Detection Filter
BilateralFilter
De-noising Filter


?
9
Our process

• a controls how much
• 1 to preserve pix

? variable

• width scale ? 2 1
• normalized to scale

10
Implementation

• edge filter de-noising filter(Laplacian of
  Gaussian)
• Generate convolution matrix
• Matrix transformation
• B Te ? A ? Te
• Scaling/Interpolation filter
• Generate convolution matrix
• Matrix transformation
• C Ts ? B ? Ts

11
Results
12
(No Transcript)
13
(No Transcript)
14
(No Transcript)
15
(No Transcript)
16
Public Survey