Report #1( imageProcessing )
Original Images
Saturn
Taj 
Tree
Mona 
GaussianSmoothing
|
Mona |
Threshold = 35 |
Threshold = 175 |
|
3x3 |
|
|
|
7x7 |
|
|
|
Tree |
Threshold = 55 |
Threshold = 200 |
|
3x3 |
|
|
|
7x7 |
|
|
Results:
As the thresholds increased for both images the edges became more blurred. However, using
A 3x3 operator from a low threshold to a higher threshold did not vary as much as did the 7x7
Operator. The images also seem to exhibit a darker hue after applying the gaussian smoothing
Operation.
MedianSmoothing
|
Saturn |
Taj |
|
|
3x3 |
|
|
|
5x5 |
|
|
Results:
The median filter worked with eliminating the noise without sacrificing too much of the image
Quality with a 3x3 operator. With a 5x5 operator the noise completely vanished but the
Sharpness of the original image faded noticeably.
Performance
|
3x3 |
5x5 |
7x7 |
9x9 |
|
|
Gaussian |
.11sec |
.17sec |
.22sec |
.28sec |
|
Median |
.77sec |
1.54sec |
2.42sec |
3.51sec |
Results:
These runs were done on a 166MHz, 586 ibm compatible computer. As the window of the
Operator grew the time increased in a linear fashion, as expected. I tried to increase the performance
Of the gaussian operation by implementing the incremental method of a 1D by 1D window operator.
However, the inherent nature of the weights associated with the gauss' methology prevented its use.