Method and apparatus for low depth of field image segmentation

• US 7,302,096 B2
• Filed: 04/11/2003
• Issued: 11/27/2007
• Est. Priority Date: 10/17/2002
• Status: Expired due to Fees

• Current Assignee: Seiko Epson Corporation (Seiko Group)
• Original Assignee: Seiko Epson Corporation (Seiko Group)
• Inventors: Kim, Changick
• Primary Examiner(s): Mancuso; Joseph
• Assistant Examiner(s): LIEW, ALEX KOK SOON
• Application Number:
  US10/412,128
• Publication Number:
  US 20040076335A1
• Time in Patent Office: 1,691 Days
• Field of Search: None
• US Class Current: 382/173
• CPC Class Codes: G06T 2207/20036 Morphological image processing
• CPC Class Codes: G06T 7/12 Edge-based segmentation
• CPC Class Codes: G06T 7/194 involving foreground-background segmentation
• CPC Class Codes: G06T 7/44 using image operators, e.g. filters, edge density metrics or local histograms
• CPC Class Codes: G06T 7/62 of area, perimeter, diameter or volume
• CPC Class Codes: G06V 10/28 Quantising the image, e.g. histogram thresholding for discrimination between background and foreground patterns

Abstract

A method for extracting an object of interest from an image is provided. The method initiates with defining an image feature space based upon frequency information. Then, the image feature space is filtered to smooth both focused regions and defocused regions while maintaining respective boundaries associated with the focused regions and the defocused regions. The filtered image feature space is manipulated by region merging and adaptive thresholding to extract an object-of-interest. A computer readable media, an image capture device and an image searching system are also provided.

Claims

1. A method for partitioning image data, comprising:

defining an image feature space based upon frequency information including, computing a higher order statistic (HOS), which does not include the variance, for each pixel value associated with the image feature space;

reducing a dynamic range of HOS values by down-scaling a HOS value associated with each pixel while still maintaining gray levels associated with each HOS value, and further reducing the dynamic range of the HOS values by limiting a maximum of the HOS value associated with each pixel while still maintaining gray levels associated with each HOS value;

filtering image data of the image feature space with morphological tools;

assigning a region of the filtered image feature space as an initial object of interest;

identifying a boundary of the initial object of interest of the filtered image feature space; and

determining a size of the initial object of interest relative to an image data size.

2. The method of claim 1, wherein the method operation of identifying a boundary of the initial object of interest of the filtered image feature space includes,

calculating a normalized overlap boundary representing a value indicating boundary pixels shared between the initial object of interest and a region bordering the initial object of interest;

if the value is greater than a threshold value then the method includes,

merging the region bordering the initial object of interest into the initial object of interest.

3. The method of claim 1, wherein the HOS represents a fourth order moment associated with each pixel value.

4. The method of claim 1, wherein the method operation of filtering image data of the image feature space with morphological tools includes,

applying the morphological tools to the image data of the image feature space in a manner that preserves the boundary of the object of interest.

5. The method of claim 1 wherein the method operation of filtering image data of the image feature space with morphological tools includes,

removing dark patches associated with focused regions of the image feature space;

and

removing bright patches associated with defocused region of the image feature space.

6. The method of claim 1, wherein the method operation of assigning a region of the filtered image feature space as an initial object of interest includes,

identifying regions of the image feature space associated with a substantially constant frequency level;

and

assigning a value to each of the identified regions based upon the substantially constant frequency level, wherein the region of the filtered image space associated with the initial object of interest is assigned a highest value.

7. A method of image segmentation, comprising:

generating a higher order statistic (HOS) map from image data, wherein the HOS may does not include the variance;

reducing the dynamic range of the HOS may by down-scaling a value associated with each pixel while still maintaining gray levels associated with each pixel, and further reducing the dynamic range of the HOS may by limiting a maximum of the value associated with each pixel while still maintaining gray levels associated with each pixel;

modifying the HOS map;

determining a boundary associated with a focused region of the modified HOS map; and

determining a final segmentation of the focused region based upon a size of a value associated with the focused region relative to an image data size.

8. The method of claim 7, wherein the method operation of determining a boundary associated with a focused region of the modified HOS map includes;

determining a value indicating a shared boundary amount between the focused region and a bordering region; and

if the value is greater than a threshold value, then the method includes,

merging the focused region and the bordering region.

9. The method of claim 7, wherein the method operation of determining a final segmentation of the focused region based upon a size of a value associated with the focused region relative to an image data size includes,

decreasing a threshold value until the size of the value associated with the focused region relative to the image data size becomes greater than about 20%.

10. The method of claim 7, wherein the focused region is an object of interest.

11. The method of claim 7, wherein the method operation of modifying the HOS map includes,

applying reconstruction by erosion to pixel values associated with the HOS map;

and

applying reconstruction by dilation to the pixel values, the applying reconstruction by dilation including,

restoring contours of components associated with the HOS map.

12. The method of claim 7, wherein the method operation of determining a final segmentation of the focused region based upon a size of a value associated with the focused region relative to an image data size includes,

defining a threshold value;

determining the size of the value associated with the focused region at the threshold value; and

reducing the threshold value until the size of the value associated with the focused region is greater than about 20% the image data size.

13. The method of claim 7, further comprising:

identifying an interior of a homogenous region of the modified HOS map; and

assigning a marker to the homogenous region.

14. A method for extracting an object of interest from an image, comprising:

defining an image feature space based upon frequency information;

calculating a higher order statistic (HOS) value for each pixel value of the image, wherein the HOS value does not include the variance;

defining a HOS map from the calculated HOS values;

reducing the dynamic range of the HOS may by down-scaling a value associated with each pixel while still maintaining gray levels associated with each pixel, and further reducing the dynamic range of the HOS may by limiting a maximum of the value associated with each pixel while still maintaining gray levels associated with each pixel; and

filtering the image feature space to smooth both focused regions and defocused regions while maintaining respective boundaries associated with the focused regions and the defocused regions.

15. The method of claim 14, further comprising:

merging similar focused regions to define an object of interest (OOI); and

determining a size of the OOI relative to the image feature space.

16. The method of claim 14, wherein the method operation of filtering the image feature space to smooth both focused regions and defocused regions while maintaining respective boundaries associated with the focused regions and the defocused regions includes,

applying a morphological filter by reconstruction to the image feature space.

17. A computer readable medium having program instructions for image segmentation, comprising:

program instructions for generating a higher order statistic (HOS) map from image data, wherein the HOS may does not include the variance;

program instructions for modifying the HOS map;

program instructions for reducing the dynamic range of the HOS may by down-scaling a value associated with each pixel while still maintaining gray levels associated with each pixel, and further reducing the dynamic range of the HOS map by limiting a maximum of the value associated with each pixel while still maintaining gray levels associated with each pixel;

program instructions for determining a boundary associated with a focused region of the modified HOS map; and

program instructions for determining a final segmentation of the focused region based upon a size of a value associated with the focused region relative to an image data size.

18. The computer readable medium of claim 17, wherein the program instructions for generating a higher order statistic (HOS) map from image data includes,

program instructions for scaling value associated with each pixel;

and

program instructions for limiting a maximum of the value associated with each pixel.

19. The computer readable medium of claim 17, wherein the program instructions for determining a boundary associated with a focused region of the modified HOS map includes;

program instructions for determining a value indicating a shared boundary amount between the focused region and a bordering region; and

program instructions for merging the focused region and the bordering region if the value is greater than a threshold value.

20. The computer readable medium of claim 17, wherein the program instructions for determining a final segmentation of the focused region based upon a size of a value associated with the focused region relative to an image data size includes,

program instructions for decreasing a threshold value until the size of the value associated with the focused region relative to the image data size becomes greater than 20%.

21. The computer readable medium of claim 17, wherein the program instructions for modifying the HOS map includes,

program instructions for applying reconstruction by erosion to pixel values associated with the HOS map;

program instructions for applying reconstruction by dilation to the pixel values, the program instructions for applying reconstruction by dilation including,

program instructions for restoring contours of components associated with the HOS map.

22. The computer readable medium of claim 17, further comprising:

program instructions for identifying an interior of a homogenous region of the modified HOS map; and

program instructions for assigning a marker to the homogenous region.