Specifications

Intel
®
Image Processing Library Reference Manual
2-8
2
Execution Architecture
Handling Overflow and Underflow
Overflow and underflow are handled in each image processing function.
The Image Processing Library uses saturation to prevent the pixel values
from potential overflow or underflow. Thus, when an overflow of a pixel
value is about to happen, this value is clamped to the maximum permissible
value (for example, 255 for an unsigned byte). Similarly, when underflow
of a value is about to happen, it is clamped to the minimum permissible
value, which is always zero for the case of unsigned bytes.
In-Place and Out-of-Place Operations
Image processing operations in the library can be in-place or out-of-place
operations. With an in-place operation, the output image is one of the input
images modified (that is, the pointer to the output image is the same as the
pointer to the input one). With an out-of-place operation, the output image
is a new image, not the same as any of the input images. Not all functions
can perform in-place operations. See Appendix A to check if a partucular
function supports in-place operation.
Image Tiling
Tiling is a method of image representation in which the image is broken up
into smaller images, or tiles, to allow for complicated memory
management schemes. Conceptually, the whole image would be
reconstructed by arranging the individual tiles in a grid. But the intent of
the tiling mechanism is to allow only a few of these tiles within an image to
reside in memory at one time. The application provides an actual memory
location for a tile only when requested to do so.
Most functions can use tiled images in the same way as non-tiled, and
procuce the same results. However, there are some differences, particularly
in the call-back requirement (see “Call-backs“ for more information).