1 / 14

Optical Character Recognition on GPU

Optical Character Recognition on GPU. Lan Vu, Spring 2012 Department of Computer Science & Engineering University of Colorado Denver. Optical Character Recognition (OCR). Convert image into text. OCR Applications. Check reader. Form reader. Mail sorter. Passport reader.

xylia
Download Presentation

Optical Character Recognition on GPU

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Optical Character Recognition on GPU Lan Vu, Spring 2012 Department of Computer Science & Engineering University of Colorado Denver

  2. Optical Character Recognition (OCR) Convert image into text

  3. OCR Applications Check reader • Form reader • Mail sorter • Passport reader

  4. How does it work ?

  5. How does it work ?

  6. Techniques for OCR Popular methods A set of algorithms to: • Transform to binary images • Reduce noise • Improve image quality • Correct skew errors • Letter segmentation • Separate text from bkground • Normalize character size • Artificial Neural Network • Hidden Markov Model • Support Vector Machine • Genetic Algorithm • Bayesian Network • Gaussian Model • K-nearest neighbor • Hybrid methods

  7. They are all computational intensive Popular methods A set of algorithms to: • Transform to binary images • Reduce noise • Improve image quality • Correct skew errors • Letter segmentation • Separate text from bkground • Normalize character size • Artificial Neural Network • Hidden Markov Model • Support Vector Machine • Genetic Algorithm • Bayesian Network • Gaussian Model • K-nearest neighborhood • Hybrid methods

  8. Image Processing with GPU • Images are just matrices of colors • RGB  grayscale  binary • Image processing algorithms • measure and adjust matrices for a purpose • Ideal for GPU • Computational intensive • Process huge amount of data • Highly data parallel computation • Minimal dependency between data elements

  9. Transform to binary images • RGB  grayscale  binary • Load image into memory as a matrix • For each cell of matrix • Grayscale_color = (R + G + B) /3 • IfGrayscale_color > K • Binary_color = Black • Else • Binary_color = White Predefined threshold e.g. K = 60

  10. Let use CUDA to run it on GPU 1378x1786

  11. Experiment of CPU vs. GPU • Test Cases • Run sequential on Intel Core 2 CPU 1.8Ghz • Run parallel on Intel Core 2 CPU 1.8Ghz (OpenMP) • Run parallel on Gefore 9400 GT Nvidia GPU(CUDA) • Test images: Bitmap 200x300 - Iterations: 10000 • Source Code: I can give a copy if you interest

  12. Experiment results

  13. You should try it on Hydra ! VS.

  14. My current plan with GPU • Sequence Alignment on GPU

More Related