Answers to questions are in answers.pdf
This library fills holes in images. Steps:
- Convert the source image to a grayscale format by taking the mean value of (r,g,b) values.
- Normalises color values from range [0-255] to [0.0-1.0]
- Assign -1 value to every pixel where defined as hole in the mask
- Find the the boundary of the hole (for either 8 or 4 connectivity)
- Applies the algorithm for every pixel in the hole
- Save the filled image in source image directory
NOTE: Mask image is the same height and width of the source image. A pixel is defined as 'hole pixel' iff its color value in the mask satisfies value > 0.5
./gradlew run --args='[image path] [mask path] [z] [epsilon] [pixel connectivity: 4/8]'
- Set z in range 2 <= z <= 25
- Set e < 1e-4
./gradlew run --args='./images/tests/test1.jpg ./images/tests/test1_mask.jpg 2 0.00001 8'
source and mask
output
Finding the boundary of the hole was a quite a challenge for me compare to other actions. I tried to implement is as simple, efficient and readable as possible. I have built quite a few unit tests for this function using jUnit. To run all tests:
./gradlew test --tests holefill.ImageProcessTest
I have included some test images in /images/tests. To run:
./gradlew run --args='./images/tests/test1.jpg ./images/tests/test1_mask.jpg 2 0.00001 8'
./gradlew run --args='./images/tests/test2.jpg ./images/tests/test2_mask.jpg 2 0.00001 4'
./gradlew run --args='./images/tests/test3.jpg ./images/tests/test3_mask.jpg 27 0.00001 8'
./gradlew run --args='./images/tests/test4.jpg ./images/tests/test4_mask.jpg 2 0.00001 4'
./gradlew run --args='./images/tests/test5.jpg ./images/tests/test5_mask.jpg 24 0.00001 8'
./gradlew run --args='./images/tests/test6.jpg ./images/tests/test6_mask.jpg 3 0.00001 8'
- Gradle Wrapper - Dependency Management and Build Tool
- JUnit 4 - testing framework for the Java
Ofir Abramovich