A segmentation process is constituted of at least one, maybe several, pipelines. A pipeline is a series of image processing operations that are executed sequentially on the input image of the pipeline, generating the output image of this pipeline. The reason for such a fragmentation of the segmentation process is, that there are image processing operations (e.g. reconstruction and impose minima) which need an input image to operate. It is possible to choose the output image of any pipeline, previous to the one the operation is part of, as an input image for this operation. An example for the use of several process pipelines is the predefined watershed segmentation process.
The input image of the first pipeline is always one of the rawdata images, either a brightfield or a fluorescent image. The output image of the last pipeline has to be an indexed segmented image. Image processing operations generating such an image are the watershed transformation and the label operation, so the last image processing operation of the last pipeline is usually one of these operations. Examples for predefined segmentation processes can be found here.
Process pipeline list
In this list the pipelines of the current process are displayed and can be selected by mouse-click. The image processing operations of this pipeline are then displayed in the operations list below.
With this button a new empty pipeline is added to the segmentation process. The new pipeline will be added behind the currently selected pipeline.
By pressing this button it is possible to remove the currently selected pipeline, with the exception if there is only one pipeline left.
In this list all image processing operations of the current process pipeline are displayed and can be selected by mouse-click. The current selected processing state can be seen in screen 1 of the main window. A selected operation can be modified using the buttons next to this list.
The first operation called 'Source image' is a special one, because it is not possible to remove it. By editing it the source image of the current process pipeline can be modified.
A new image processing operation is added behind the currently selected one. Therefor a new window to add operations is opened.
By pressing this button it is possible to modify the parameters of the currently selected operation or to replace it with a different on. For this purpose a window to edit operations is opened, similiar to the one for adding operations.
The currently selected operation is removed from the list.
All image processing operations of the current process pipeline are removed at once.
By pressing preview the segmentation process is executed for the current image. That means that all image processing operations are executed to get the final image of the last pipeline (an indexed segmented image), and the properties of the cells are determined from this image. As a result the recognized cells are displayed in screen 2 of the main window, marked by blue frames on the original image.
The results which can be seen here are only temporary and are not saved. To finally execute the segmentation process for the current image or a series of images use the 'Execute/Execute all'-button in the Segmentation process panel of the main window.
Here, thumbnails for the currently loaded segmentation process are displayed. These thumbnails show the first and the last image of each process pipeline. They are generated when a process is saved and displayed everytime it is reloaded.
An existing segmentation process can be inspected by starting at the first operation of the first process pipeline and then proceeding to the subsequent operations. Process pipelines and image processing operations can be selected by clicking on them. The current state of the segmentation process for the currently selected image can be seen on screen 1 of the main window. If an image processing operation doesn't have the desired effect it can be modified using the 'Edit'-button. There are several predefined segmentation processes which can serve as examples and which be inspected and modified.
If a new segmentation process shall be created start by editing the source image operation and select which image shall be used as the source image. This can be either the brightfield or the fluorescence image. Afterwards add image processing operations using the 'Add'-button to achieve a segmentation of the source image resulting in a binary image where every connected region corresponds to one cell in the original image. The last step is to label the segmented image. Usually it is a good idea to do this in a new process pipeline, which can be added using the 'Add'-button in the pipeline-panel. Use the final image of the first pipeline as the source image for the new one and add the label operation to it. This partitioning of the segmentation process can be advantageous if the image has to be corrected manually. The best place to do this is before the labeling, at the end of the first pipeline.