Image Processing

Collaboration diagram for Image Processing:

Modules
	Image Statistics
	Image Analysis
	Other Domain Transform Operations
	Fourier Transform Operations
	OpenMP Utilities
	Image Resize
	Geometric Operations
	Morphology Operations for Gray Images
	Morphology Operations for Binary Images
	Binary Image Operations
	Rank Convolution Operations
	Convolution Operations
	Point Based Custom Operations
	Arithmetic Operations
	Additional Image Quantization Operations
	Histogram Based Operations
	Color Processing Operations
	Logical Arithmetic Operations
	Synthetic Image Render
	Tone Gamut Operations
	Threshold Operations
	Special Effects
	Remote Sensing Operations
	Image Conversion
	Image Enhance Utilities in Lua

---

Detailed Description

Several image processing functions based on the imImage structure.

You must link the application with "im_process.lib/.a/.so". In Lua call require"imlua_process".
Some complex operations use the Counter.
There is no check on the input/output image properties, check each function documentation before using it.

To enable OpenMP support use the "im_process_omp.lib/.a/.so" libraries. In Lua call require"imlua_process_omp".
Notice that multi-threading can be slower than single thread because of the overhead introduced by the threads configuration.
When using the "im_process_omp" library you can reduce that overhead by using the imProcessOpenMPSetMinCount and imProcessOpenMPSetNumThreads functions. But notice that this is not the same thing as using the library without support for OpenMP.

The parallelization in im_process involves only loops, usually for all the pixels in the image. To accomplish that we had to first isolate the Counter code, so the counting could also be done in parallel. Then we made sure that all loops contain only local variables to avoid unnecessary shared variables that could lead to incorrect results. In a few places we use the "atomic" directive to be able to compute histograms and other counts. But min/max computation must be done in single thread because of limitations in OpenMP support in C (in Fortran it would be easy to implement).

For more information on OpenMP:
http://www.openmp.org