 |
ImFusion SDK 4.3
|
|
ImFusion SDK 4.3
|
The ImFusionLib distinguishes between two pixel value representations: original and stored. The original pixel values are the values as they are stored in the original source. For instance, in CT images originating from DICOM this could be Hounsfield Units. The stored pixel values, on the other hand, are values as they are stored in memory.
The relationship between the two representations is defined by the Image::scale and Image::shift properties. The following conversion rules apply:
Original pixel values should always be used for display purposes (e.g. HU values in a histogram) while most algorithm will use storage values internally for their computations.
Be aware that when using ApplicationController or MainWindowBase they will register a prepare-data-callback that will automatically convert signed images added to the DataModel to unsigned images by applying a shift with the negative minimum of the original values (see also MemImage::prepare). This is the default behavior of the ImFusion Suite and allows for a wider range of supported algorithms as they don't need to deal with signed data. Doing so in custom applications is optional. Keep in mind, that you should not assume original and storage values to be the same.
If not explicitly noted otherwise, all function arguments regarding pixel values expect storage values.
When uploading images to OpenGL textures (i.e. GlImage) OpenGL will perform an implicit conversion to a normalized floating point value domain since texture access in GLSL shaders always returns float. Integral image types (both signed and unsigned) are mapped to a [0..1] range. Float image types keep their domain.
When using ImFusion SDK primitives to convert between different image representations (e.g. MemImage and GlImage) the correct conversion of value domains is performed automatically. If you need to convert between different pixel value domains manually, you can use the following helper functions: