Methods for adaptive and progressive gradient-based multi-resolution color image segmentation and systems thereof

• US 8,515,171 B2
• Filed: 01/11/2010
• Issued: 08/20/2013
• Est. Priority Date: 01/09/2009
• Status: Expired due to Fees

• Current Assignee: Hewlett-Packard Development Company, L.P. (HP Inc.), Rochester Institute of Technology
• Original Assignee: Hewlett-Packard Development Company, L.P. (HP Inc.), Rochester Institute of Technology
• Inventors: Vantaram, Sreenath Rao, Saber, Eli, Dianat, Sohail, Shaw, Mark, Bhaskar, Ranjit
• Primary Examiner(s): Alavi, Amir
• Assistant Examiner(s): CESE, KENNY A
• Application Number:
  US12/655,933
• Publication Number:
  US 20100183225A1
• Time in Patent Office: 1,317 Days
• Field of Search: None
• US Class Current: 382/173
• CPC Class Codes: G06T 2207/10024 Color image
• CPC Class Codes: G06T 2207/20036 Morphological image processing
• CPC Class Codes: G06T 2207/20064 Wavelet transform [DWT]
• CPC Class Codes: G06T 7/11 Region-based segmentation
• CPC Class Codes: G06T 7/136 involving thresholding
• CPC Class Codes: G06T 7/155 involving morphological operators
• CPC Class Codes: G06T 7/168 involving transform domain methods
• CPC Class Codes: G06T 7/174 involving the use of two or more images
• CPC Class Codes: G06T 7/187 involving region growing; involving region merging; involving connected component labelling

Abstract

A multi-resolution color image segmentation algorithm which takes advantage of gradient information in an adaptive and progressive framework is described. A gradient-based segmentation method is initiated with a dyadic wavelet decomposition scheme of an arbitrary input image, accompanied by a vector gradient calculation of its color converted counterpart. The resultant gradient map is used to automatically and adaptively generate thresholds for segregating regions of varying gradient densities, at different resolution levels of the input image pyramid. In combination with a confidence map and non-linear spatial filtering techniques, regions of high confidence are passed from one resolution level to another until the final segmentation at highest (original) resolution is achieved.

Claims

1. A method for image segmentation, the method comprising:

performing by an image processing computing device a dyadic wavelet decomposition of an input image with automatic determination of a number of decomposition levels;

adaptively generating by the image processing computing device gradient thresholds for initial clustering and region processing at different resolution levels of the input image from a gradient histogram obtained from the performed dyadic wavelet decomposition;

implementing by the image processing computing device progressively thresholded multi-resolution region growth on the initial clustering and dyadic wavelet decomposition of the input image;

identifying by the image processing computing device transferable regions for multi-resolution information transfer, based on the implemented progressively thresholded multi-resolution region growth and a histogram analysis of gradient information; and

merging by the image processing computing device the identified regions to provide interim results at arbitrary low resolution levels and a final segmentation map at a dyadic scale equal to that of the input image.

2. The method as set forth in claim 1 further comprising converting by the image processing computing device the input image from RGB to CIE L*a*b* before the adaptively generating thresholds.

3. The method as set forth in claim 1 wherein the adaptively generating gradient thresholds further comprises:

determining by the image processing computing device a vector gradient calculation on the input image to generate a resultant edge map; and

using by the image processing computing device the resultant edge map to adaptively generate the gradient thresholds for the initial clustering and the region processing at different resolution levels, of the input image.

4. The method as set forth in claim 1 wherein the performing the dyadic wavelet decomposition of the input image further comprises classifying by the image processing computing device regions of varying gradient densities at different levels.

5. The method as set forth in claim 1 wherein the implementing progressively thresholded multi-resolution region growth utilizes distributed dynamic seed addition.

6. The method as set forth in claim 1 further comprising performing by the image processing computing device texture modeling on the initial clustering and dyadic wavelet decomposition of the input image, the merging is further based on the texture modeling.

7. A non-transitory computer readable medium having stored thereon instructions for image segmentation comprising machine executable code which when executed by at least one processor, causes the processor to perform steps comprising:

performing a dyadic wavelet decomposition of an input image with automatic determination of number of decomposition levels;

adaptively generating gradient thresholds for initial clustering and region processing at different resolution levels of the input image from a gradient histogram obtained from the performed dyadic wavelet decomposition;

implementing progressively thresholded multi-resolution region growth on the initial clustering and dyadic wavelet decomposition of the input image;

identifying transferable regions for multi-resolution information transfer, based on the implemented progressively thresholded multi-resolution region growth and a histogram analysis of gradient information; and

merging the identified regions to provide interim results at arbitrary low resolution levels and a final segmentation map at a dyadic scale equal to that of the input image.

8. The medium as set forth in claim 7 further comprising converting the input image from RGB to CIE L*a*b* before the adaptively generating thresholds.

9. The medium as set forth in claim 7 wherein the adaptively generating gradient thresholds further comprises:

determining a vector gradient calculation on the input image to generate a resultant edge map; and

using the resultant edge map to adaptively generate the gradient thresholds for the initial clustering and the region processing at different resolution levels, of the input image.

10. The medium as set forth in claim 7 wherein the performing the dyadic wavelet decomposition of the input image further comprises classifying regions of varying gradient densities at different levels.

11. The medium as set forth in claim 7 wherein the implementing progressively thresholded multi-resolution region growth utilizes distributed dynamic seed addition.

12. The medium as set forth in claim 7 further comprising performing texture modeling on the initial clustering and dyadic wavelet decomposition of the input image, the merging is further based on the texture modeling.

13. An image processing computing apparatus comprising a memory coupled to a processor configured to execute programmed instructions stored in the memory including instructions for implementing:

performing a dyadic wavelet decomposition of an input image with automatic determination of number of decomposition levels;

adaptively generating gradient thresholds for initial clustering and region processing at different resolution levels of the input image from a gradient histogram obtained from the performed dyadic wavelet decomposition;

implementing progressively thresholded multi-resolution region growth on the initial clustering and dyadic wavelet decomposition of the input image;

identifying transferable regions for multi-resolution information transfer, based on the implemented progressively thresholded multi-resolution region growth and a histogram analysis of gradient information; and

merging the identified regions to provide interim results at arbitrary low resolution levels and a final segmentation map at a dyadic scale equal to that of the input image.

14. The apparatus as set forth in claim 13 wherein the processor is further configured to execute programmed instructions stored in the memory further comprising converting the input image from RGB to CIE L*a*b* before the adaptively generating thresholds.

15. The apparatus as set forth in claim 13 wherein the processor is further configured to execute programmed instructions stored in the memory for the adaptively generating gradient thresholds further comprises:

determining a vector gradient calculation on the input image to generate a resultant edge map; and

using the resultant edge map to adaptively generate the gradient thresholds for the initial clustering and the region processing at different resolution levels, of the input image.

16. The apparatus as set forth in claim 13 wherein the processor is further configured to execute programmed instructions stored in the memory for the performing the dyadic wavelet decomposition of the input image further comprises classifying regions of varying gradient densities at different levels.

17. The apparatus as set forth in claim 13 wherein the implementing progressively thresholded multi-resolution region growth utilizes distributed dynamic seed addition.

18. The apparatus as set forth in claim 13 wherein the processor is further configured to execute programmed instructions stored in the memory further comprising performing texture modeling on the initial clustering and dyadic wavelet decomposition of the input image, the merging is further based on the texture modeling.