Method and apparatus for re-classifying color image pixels classified by single channel segmentation
First Claim
1. An apparatus for single channel segmentation of color images, the apparatus comprising:
- a data reduction circuit for converting a first set of pixels of a multi-channel color input image to a second set of pixels of a single channel representation of said multi-channel color input image;
a single channel segmentation circuit operative to classify said second set of pixels into a plurality of segmentation classes by generating first classification data representative of said plurality of segmentation classes and associating said first classification data with each of said second set of pixels; and
, a re-classification circuit receiving said first classification data and said first set of pixels of said multi-channel color input image, the re-classification circuit being adapted to selectively re-classify ones of said second set of pixels into an “
other”
segmentation class by generating second classification data representative of said “
other”
segmentation class and associating said second classification data with said ones of said second set of pixels.
7 Assignments
0 Petitions
Accused Products
Abstract
A method and apparatus are provided for use in single channel segmentation of color images for reclassifying pixels which were inappropriately classified as belonging to a “black” or “white” class. The apparatus includes a re-classification circuit receiving first classification data generated by a single channel segmentation circuit operative to classify pixels of a multi-color channel input image. The re-classification circuit selectively re-classifies pixels that were previously classified by the single channel segmentation circuit into an “other” segmentation class based on a comparison of classification data associated with the pixels relative to a set of predefined classification data types. Particularly, the re-classification circuit reclassifies selected ones of the pixels into the “other” segmentation class when first classification data generated by the single channel segmentation circuit labels the pixels as a “white” or “black” data type.
-
Citations
13 Claims
-
1. An apparatus for single channel segmentation of color images, the apparatus comprising:
-
a data reduction circuit for converting a first set of pixels of a multi-channel color input image to a second set of pixels of a single channel representation of said multi-channel color input image;
a single channel segmentation circuit operative to classify said second set of pixels into a plurality of segmentation classes by generating first classification data representative of said plurality of segmentation classes and associating said first classification data with each of said second set of pixels; and
,a re-classification circuit receiving said first classification data and said first set of pixels of said multi-channel color input image, the re-classification circuit being adapted to selectively re-classify ones of said second set of pixels into an “
other”
segmentation class by generating second classification data representative of said “
other”
segmentation class and associating said second classification data with said ones of said second set of pixels.- View Dependent Claims (2, 3, 4, 5, 6, 7, 8)
said single channel segmentation circuit is operative i) in a non-failure mode to classify said second set of pixels into a plurality of non-failure segmentation classes by generating said first classification data representative of said plurality of non-failure segmentation classes and associating said first classification data with each of said second set of pixels, and ii) in a failure mode to classify said second set of pixels into a plurality of failure segmentation classes by generating said first classification data representative of said plurality of failure segmentation classes and associating said first classification data with each of said second set of pixels; and
,said re-classification circuit is adapted to selectively re-classify said ones of said second set of pixels associated with said first classification data representative of said plurality of failure segmentation classes into said “
other”
segmentation class.
-
-
5. The apparatus according to claim 4 wherein:
-
said single channel segmentation circuit is operative in said failure mode to classify said second set of pixels into a plurality of black/white background segmentation classes; and
,said re-classification circuit is adapted to selectively re-classify said ones of said second set of pixels associated with said first classification data representative of said black/white background classes into said “
other”
segmentation class based upon a comparison of said first set of pixels of said multi-channel color input image with a set of predetermined threshold values.
-
-
6. The apparatus according to claim 5 wherein said re-classification circuit is adapted to selectively re-classify said ones of said second set of pixels into an “
- other_light”
segmentation class and an “
other_dark”
segmentation class based upon a comparison between said first set of pixels of said multi-channel color input image with a set of predetermined threshold values.
- other_light”
-
7. The apparatus according to claim 2 wherein:
-
said single channel segmentation circuit is operative to generate a confidence value signal together with said first classification data, the confidence value signal being representative of a relative certainty of said second set of pixels belonging to said plurality of segmentation classes; and
,said re-classification circuit is responsive to said confidence value signal to selectively re-classify second ones of said second set of pixels into said “
other”
segmentation class by generating said second classification data representative of said “
other”
segmentation class and associating said second classification data with said second ones of second set of pixels.
-
-
8. The apparatus according to claim 7 wherein said re-classification circuit is responsive to a level of said confidence value signal relative to a predetermined threshold to selectively re-classify said second ones of said second set of pixels into said “
- other”
segmentation class.
- other”
-
9. A method of re-classifying color image pixels previously classified by single channel segmentation, the method comprising the steps of:
-
converting a first set of pixels of a multi-channel color input image to a second set of pixels of a single channel representation of said multi-channel color input image;
classifying said second set of pixels into a plurality of segmentation classes by generating first classification data representative of said plurality of segmentation classes and associating said first classification data with each of said second set of pixels; and
,based on a comparison of said first classification data associated with said ones of said second set of pixels relative to a set of predetermined classification data types, selectively re-classifying ones of said second set of pixels into an “
other”
segmentation class by generating second classification data representative of said “
other”
segmentation class and associating said second classification data with said ones of said second set of pixels.- View Dependent Claims (10, 11, 12, 13)
said step of classifying said second set of pixels includes i) classifying said second set of pixels into a plurality of non-failure segmentation classes by generating said first classification data representative of said plurality of non-failure segmentation classes and associating said first classification data with each of said second set of pixels, and ii) classifying said second set of pixels into a plurality of failure segmentation classes by generating said first classification data representative of said plurality of failure segmentation classes and associating said first classification data with each of said second set of pixels; and
,the step of re-classifying said ones of said second set of pixels includes re-classifying said ones of said second set of pixels associated with said first classification data representative of said plurality of failure segmentation classes into said “
other”
segmentation class.
-
-
12. The method according to claim 11 wherein:
-
the step of classifying said second set of pixels includes classifying said second set of pixels in said failure mode into a plurality of black/white background segmentation classes; and
,the step of re-classifying said ones of said second set of pixels includes re-classifying said ones of said second set of pixels associated with said first classification data representative of said black/white background classes into said “
other”
segmentation class based upon a comparison of said first set of pixels of said multi-channel color input image with a set of predetermined threshold values.
-
-
13. The method according to claim 12 wherein the step of re-classifying said ones of said second set of pixels includes re-classifying said ones of said second set of pixels into an “
- other_light”
segmentation class and an “
other_dark”
segmentation class based upon a comparison between said first set of pixels of said multi-channel color input image with a set of predetermined threshold values.
- other_light”
Specification