4.4. Global Image Processing Concepts

4.4. Global Image Processing Concepts
	Chapter 4. Image Processing Concepts

4.4.1. Temporary Global Concept

This concept has been designed for algorithms that require a single, very efficient random image access to calculate a result.

	Important
If possible, try to avoid this approach! It supports only limited image sizes which depend on the available memory!

Important

If possible, try to avoid this approach!

It supports only limited image sizes which depend on the available memory!

Procedure:

Read entire input image (e.g., when first voxel or page is requested or when the user starts the algorithm by pressing some button)
Analyze the image and save all results
Free allocated image buffer
For page requests, the results are used to fill those pages

Advantages:

Global algorithms are easy to implement, because all image data is directly available.
Uses the memory only temporarily, i.e the memory is available again when the process has been finished.

Disadvantages:

Results must be saved/buffered in data structures and must be reconverted into requested pages. Hence, additional structures are required.
The image might not fit into memory.
Even if only one voxel or a small page is requested, the entire volume must be processed.
Danger of heap fragmentation.

4.4.2. Global Image Processing Concept

This concept is needed for time-critical random image access to calculate a result.

	Important
If possible, try to avoid this approach! It supports only limited image sizes which depend on the available memory!

Important

If possible, try to avoid this approach!

It supports only limited image sizes which depend on the available memory!

Procedure:

Read entire input image (e.g., when first voxel or page is requested or when the user starts the algorithm by pressing some button)
Analyze the image and save all results
Free allocated image buffer
For page requests, the results are used to fill those pages.

Advantages:

Easy and fast implementation.

Disadvantages:

Degenerates the page-based, cached calculation process.
Fails on images which do not fit into memory.
Must map the entire image and blocks large memory areas for a long time. It cannot/should not be used in larger networks or applications because the "Out of Memory" state is easily reached.
Danger of heap fragmentation.
Usually the image must be passed page-based to the output which also requires additional implementation of page-based access to image.

Examples:

See Section 4.2.3, “Slice-Based Concept”: The requested page has the size of the entire image.
See Section 4.3.1, “Random Access Concept (Tile Requesting)”: The entire image is requested as a tile.

4.4.3. `BitImage` Concept

In the page-based image processing concept of the ML, Boolean data types are not available (nor are they planned).

The BitImage class can be used as an alternative option.

Advantages:

Easy-to-use and compact image
quite compact image although it is a global image
relatively fast.

Disadvantages:

Not paged, i.e., global; this, however, is not really problematic because only bits are stored

See Section 2.3.6, “ BitImage ” for a detailed overview.

4.4.4. `MemoryImage` Concept

Algorithms that need access to a whole non-paged memory-mapped image might use the MemoryImage approach for image processing (see Section 2.3.8, “ MemoryImage ”). Note that this breaks the page-based approach - nevertheless it is supported by the ML. It is integrated as a special member of the PagedImage in such a way that it can be handled in parallel or instead of a paged image (see Section 2.3.4, “ PagedImage ” ).


4.3. Concepts for Partially Global Image Processing		4.5. Miscellaneous Modules