×

Detecting objects of interest in still images

  • US 8,401,250 B2
  • Filed: 04/05/2010
  • Issued: 03/19/2013
  • Est. Priority Date: 02/19/2010
  • Status: Active Grant
First Claim
Patent Images

1. A computer implemented method for detecting interest sections in a still image, comprising:

  • segmenting one or more sub images of said still image based on pixels of interest in one or more color spaces, said one or more sub images being categorized into one or more of interest sub images and a plurality of non interest sub images based on said segmentation;

    matching one of a gray scale version of said interest sub images and a binary image version of said interest sub images with a predefined template for filtering said interest sub images based on relative densities of said pixels of interest and relative average intensities of predetermined features in said interest sub images;

    determining a plurality of prospective image sections comprising one or more of a plurality of prospective interest sections and a plurality of prospective near interest sections by performing discriminative feature analyses of said filtered interest sub images, wherein said discriminative feature analyses are processed by a boosted cascade of classifiers; and

    detecting said interest sections in said still image from said prospective interest sections and said prospective near interest sections by said boosted cascade of classifiers.

View all claims
  • 2 Assignments
Timeline View
Assignment View
    ×
    ×