Institutional Repository of Vilnius University
Hofo transformacijos taikymas mikroskopijos vaizdų analizėje /
Abstract
dc:descriptionMicroscopy image analysis is a broad subject, therefore in this work we chose to analyse defects in phospholipid membrane which are seen in atomic force microscope images. It is important field of study because these defects damage cells and can cause diseases for both human and animals. Since lesions on membrane are circular shaped we chose to use Hough transform for circle recognition in images. We also modified the algorithm by adapting the main idea of probabilistic Hough transform, which is random sampling. Then we calculated average precision and measured compilation speed for each image in database. Also we noticed that Hough transform algorithm is largely dependent on parameters given, hence we did some experiments to find out which parameters provide the most accuracy when recognizing defects in given images. After the research we came to the conclusion that average precision is highest when canny edge detection algorithm has set Gaussian filter to 2, low hysteresis threshold to 10, high hysteresis threshold to 30, also it is crucial to set minimum distance between center point of circles and skimage.transform.hough_circle_peaks function provides us with possibility to do so, accuracy of detections are the highest when distance is set to 10 for both x and y axis. One of the major drawbacks of the Hough transform in that the complexity and speed of the method is largely dependent on the size of input data. Our modified Hough transform cuts compilation speed nearly in half compared to traditional Hough tranform, but average precision of detections drops significantly. We have also noticed that the simpler the image, the higher the average precision. Highest accuracy is received when defects in image do not form clusters, overlap with each other or form only a circular arc rather than a full circle.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2021
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Bikulčiūtė, Sigrida,
- Contributors dc:contributor
-
- Raila, Tomas
Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- lit
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD88379092&prefLang=en_US