If a simple rectangle is not suitable to enclose a region of interest (ROI) or to split the image for specialised image segmentations, here is a quick guide on how to solve this problem in LemnaGrid.
To define multiple polygonial ROIs efficiently, we can
- create an image mask on the basis of a reference image
- refine the mask with a raster graphics editor, e.g. Gimp, or Paint.NET
- apply the image mask in LemnaGrid with Fixed data
Fixed data has two features:
- stores any input data, including binary image (right clik on the device – set fixed data)
- returns stored data as output (connect output from Fixed data to input of LemnaGrid device)
- Load a reference image, e.g. with Disk universal reader
- Pick white as foreground with Foreground separation
- refine the binary image with morphological operators:
- Fill areas to remove pixel artefacts or to close holes
- Erosion/Dilation to adjust the size of the foreground shapes.
- Save the black and white output image as .png file on the disk.
- refine the exported image-mask.png with an image graphic tool if necessary
- import the image-mask-refined.png file into LemnaGrid with Disk universal reader
- convert the type to binary image
- integrate the binary mask into LemnaGrid pipeline with Fixed data
- disconnect and remove Fixed data from (Disk universal reader – Foreground Separation)
Here is how to use Fixed data for VIS side view 120 degree image segmentation
Notice the position of the car in 120 degree
Here is an application for VIS top image segmentation