You can use morphological operations like closing. Choose the structural element you use based on the geometry of the lines you wish to connect
Dear Muhammad,
Are these lines the output of an edge detection filter or are they boundaries in the original images between two relatively homogeneous regions?
Do you suspect that the brokenness is due to insufficient local contrast or due to noise (sampling or instrumentation)?
Do you expect the lines to be locally straight or curved?
Antonio Lucero
Dear muhammad,
maybe image parsing can help you. you can find some of the papers in the below link:
http://www.stat.ucla.edu/~sczhu/
best,
Hamid
Morphological operations such as closing, using linear structuring elements of various orientations may be a method of choice. You can find details in the following publication:
Terol-Villalobos, I. R.; Morales-Hernández, L. A. & Herrera-Ruiz, G. Directional Structures Detection Based on Morphological Line-Segment and Orientation Functions Discrete Geometry for Computer Imagery, Springer Berlin / Heidelberg, 2008, 4992, 458-469
Otherwise, Gabor filters may also be of interest.
Article Directional morphological filtering
Conference Paper Directional Structures Detection Based on Morphological Line...
Try using function houghlines in matlab. It might help to connect the broken lines in the image.
Do you have a model or some knowledge of the image capturing process that can differentiate between 1) a "real" broken line and 2) an apparent broken line that, in reality, is not broken? What about comparing the images with the actual objects?
I think you need such truths and such models so you can optimize your image processing algorithm(s).
@Leio. To what to you attribute the brokenness to? Is it instrumentation noise? Is it due to locally low contrast? Is it due to a variegated background or a variegated object? Is it due to low spatial or intensity resolution? What is your theory?
- Badges
- Science topic
- Similar topics
- Computer Science
- Images