# NeatVision Counters

Mar 302010

### Develop a two NeatVision programs capable of automatically counting the number of lines in the grey scale image lines.gif in a robust manner. What are the advantages and disadvantages of each approach?

A. Using a threshold of 100, the image is converted to binary. The image is passed through an erode loop twice to thin the lines making it easier for the limb end detector. The limb end detector marks the ends of each line. Each line has 2 points. The white blob counter counts the limb ends. The mathsdivide operator divides the number of points by 2, giving the final result of 5, which is the correct number of lines.

B. Using a threshold of 100, the image is converted to binary. Hough transform will only work with a binary image. The Hough Transform is used to segment and detect the number of lines. Using a threshold of 100, the image is converted to binary again. The image is eroded 5 times, to remove the excess data and dilated once. Whiteblobcount is used to count the lines. # Chapter 2 – Testing

## Salt and Pepper Noise :

The maximum threshold of Salt and Pepper Noise is 0.0001.  B.

Salt and Pepper Noise :

The system is not robust to any Salt and Pepper Noise. Testing ceased at 0.000001. A.

Gaussian Noise :

The maximum threshold of Gaussian Noise is 0.8.  B.

Gaussian Noise :

The maximum threshold of Gaussian Noise is 1.3.

A.

Raleigh Noise :

The maximum threshold of Raleigh Noise is 26.  B.

Raleigh Noise :

The maximum threshold of Raleigh Noise is 17.7.  A.

Poisson Noise :

The maximum threshold of Poisson Noise is 27.  B.

Poisson Noise :

The maximum threshold of Poisson Noise is 16.0.  A.

This system is not robust to additive noise.

B.

This system is not robust to additive noise.

Robustness to Scaled Images :

The original image is 256 x 256. The height and width were increased or decreased depending on the purpose.

A.

Minumum Scaled Threshold :

The image was scaled down but would not function. The minimum threshold is the same size as the original.

B.

Minumum Scaled Threshold :

The image was scaled down but would not function. The minimum threshold is the same size as the original.

A.

Maximum Scaled Threshold :

The maximum threshold is 260 x 260  B.

Maximum Scaled Threshold :

The image was scaled up but would not function. The maximum threshold is the same size as the original.

Distorting Width

A.

Minimum Width Threshold :

The minimum threshold of the width is 254.  B.

Minimum Width Threshold :

The width of the image was decreased but would not function. The minimum threshold is the same size as the original.

A.

Maximum Width Threshold :

The maximum width is 263. B.

Maximum Width Threshold

The maximum threshold of the width is 274.  Distorting Height

A.

Maximum Height Threshold :

The maximum height was 263. B.

Maximum Height Threshold :

The height of the image was increased but would not function. The maximum threshold is the same size as the original. A.

Minimum height Threshold :

The height of the image was decreased but would not function. The minimum threshold is the same size as the original.

B.

Minimum height Threshold :

The height of the image was decreased but would not function. The minimum threshold is the same size as the original.

A.

Robustness to Image Rotation:

This system was not robust to any rotational changes. B.

Robustness to Image Rotation The system would only function at 0 and 180 degrees.

A.

Robustness to another image

Bcode2 was tried but the system failed.

B.

Robustness to another image

Bcode2 was tried but the system failed. # Bibliography

A. Machine Vision Algorithms in Java, Paul F. Whelan and Derek Molloy.

# References

1. http://www.eeng.dcu.ie/%7Ewhelanp/EE425/protected_material/, EE425 Course Notes, Colour Image Analysis

2. www.neatvision.com, NeatVision Developers Guide