featuresHOGColor.m 836 B

12345678910111213141516171819202122232425
  1. % function hogFeature = featuresHOGColor( img, sbin );
  2. %
  3. % BRIEF:
  4. % Compute an array of extended HOG features for a given color image.
  5. % Differs from the original implementation (featuresHOGorig.cc) by not
  6. % leaving a boundary of sbin pixels at the border of the image "unused".
  7. %
  8. % Advantage: k*sbin pixel result in k cells (orig: k-2)
  9. % Disadvantage: block normalization ill posed on boundary
  10. %
  11. % INPUT:
  12. % img -- (x,y,3) double array, input image
  13. % sbin -- double scalar, number of pixels each cell covers in x and y
  14. % direction
  15. %
  16. % OUTPUT:
  17. % hogFeature -- (x/sbin, y/sbin, 32) double array,
  18. % extracted hog array, last dim equals 0
  19. %
  20. % NOTE:
  21. % Don't miss to mex (compile) the .cc-file!
  22. %
  23. % author: Alexander Freytag
  24. % last update: 11-03-2014 ( dd-mm-yyyy )
  25. %