Rhonda has developed the computer vision software by request of Semiconductors Insight for automatic adjusting images obtained from an electronic microscope using Fourier quick transformation. The histograms of multiple images were adjusted based on chosen sample image so called “golden standard”.

Development environment.

• Language:
  • Fixed point Matlab for algorithm development
  • C/C++
• Compiler:
  • Matlab 7.5.0
  • Visual Studio 2005/2008
• Libraries:
  • Windows SDK
  • WinAPI
  • Boost
  • Intel® Integrated Performance Primitives

Project’s quick facts.

• Image input format: memory map 8bit, grayscale uint8
• Image output format: memory map 8bit, grayscale uint8, same size as input image. No scale/stretch conversion
• Maximum Image size: 8192x8192x8bit pixels, 64Mb
• Using DSP, GPU is not allowed, all processing is performed by CPU
• Maximum memory usage: Up to 1G per image (optimal 640Mb)
• Multi threading: Up to 2 separate cores can be used
• Operation system: Windows XP/Vista
• Performance: not real-time. Up to 30 seconds for processing 1 64Mb image
• Code size is about 1,900 LOC