EE 475 DIGITAL IMAGE PROCESSING

1
EE 475 DIGITAL IMAGE PROCESSING

• TERM PROJECT
• FINAL PRESENTATION

2
VECTOR MEDIAN FILTER APPLICATION OVER COLOR IMAGES
3
MEDIAN FILTER ON GRAYSCALE IMAGES
Original Image
4
MEDIAN FILTER ON GRAYSCALE IMAGES
Image with Gaussian Noise N(0,600)
5
MEDIAN FILTER ON GRAYSCALE IMAGES
Resulting Image after Median Filtering
6
MEDIAN FILTER ON COLOR IMAGES

• The first application of median filter is on
  three color matrices separately.

• The drawback of this method is that the separate
  elements are almost always correlated and such
  usage of median filter does not utilize this
  property.

7
MEDIAN FILTER ON COLOR IMAGES
The original image
8
MEDIAN FILTER ON COLOR IMAGES
The resulting image after applying the median
filter separately on each matrix
9
MEDIAN FILTER ON COLOR IMAGES

• The second application of median filter on color
  images is the vector median filter.
• This time, the filter is applied on each matrix
  element treated as a vector.

10
BLOCK DIAGRAM OF VECTOR MEDIAN FILTER PROCESS
RED
The filter window inputs are the same elements in
each matrix
GREEN
BLUE
Three components of a color image
11
BLOCK DIAGRAM OF VECTOR MEDIAN FILTER PROCESS
The resulting vector is replaced into the new
image
The output vector of the filter is evaluated
12
MEDIAN FILTER ON COLOR IMAGES
The Original Image
13
MEDIAN FILTER ON COLOR IMAGES
Resulting Image after applying vector median
filter
14
MEDIAN FILTER ON COLOR IMAGES WITH NOISE
Image containing Gaussian Noise of zero mean and
600 variance
15
MEDIAN FILTER ON COLOR IMAGES WITH NOISE
Resulting Image after applying vector median
filter
16
MEDIAN FILTER ON COLOR IMAGES WITH NOISE
Image containing 15 Salt and Pepper Noise
17
MEDIAN FILTER ON COLOR IMAGES WITH NOISE
Resulting Image after applying vector median
filter
18
MEDIAN FILTER ON COLOR IMAGES WITH NOISE
Image containing Gaussian Noise of zero mean and
600 variance
19
MEDIAN FILTER ON COLOR IMAGES WITH NOISE
Window size3X3
20
MEDIAN FILTER ON COLOR IMAGES WITH NOISE
Window size5X5
21
Second Approach to Evaluation and Classification
of Vectors

• Instead of calculating vector norms, i.e.
  magnitudes, angles between the vectors in the
  filter window are evaluated

• The advantage of this method is that filter
  outputs are close to each other and can be
  manipulated by 2D techniques

22
BASIC VECTOR DIRECTIONAL FILTER

• The component which has the median angle within
  the filter mask is chosen as the output of the
  filter

23
Basic Vector Directional Filter
24
Basic Vector Directional Filter
Window size3X3
25
Basic Vector Directional Filter
Window size5X5
26
Basic Vector Directional Filter
27
Basic Vector Directional Filter
Window size3X3
28
Basic Vector Directional Filter
Window size5X5
29
GENERAL VECTOR DIRECTIONAL FILTER

• Basic vector median filter is used with a
  multitude of outputs, resulting in the centermost
  directioned vectors

• 2D gray-level morphological and order statistics
  filters are applied to the resultant set

30
GENERAL VECTOR DIRECTIONAL FILTER
31
GENERAL VECTOR DIRECTIONAL FILTER
followed by

• median filter

• mean filter

• a-trimmed mean filter

• max filter

• min filter

32
General VDF followed by median filter
33
General VDF followed by median filter
34
General VDF followed by mean filter
35
General VDF followed by mean filter
36
General VDF followed by mean filter
37
General VDF followed by mean filter
38
General VDF followed by a-trimmed mean filter
39
General VDF followed by a-trimmed mean filter
40
General VDF followed by a-trimmed mean filter
41
General VDF followed by a-trimmed mean filter
42
General VDF followed by max filter
43
General VDF followed by max filter
44
General VDF followed by min filter
45
General VDF followed by min filter