Entropy filter, and area extracting method using the filter
AssigneeJapan Science and Technology Corporation
InventorsShugo Hamahashi, Shuichi Onami, Hiroaki Kitano
Life of this patent
Estimating the term…
Abstract
A smooth portion is efficiently extracted from an image. There is provided an entropy filter in which a noted pixel (x, y) is determined in an original image, said image is partitioned by a window of size (width, height)=(A, B) from the noted pixel, entropy of the partitioned window is calculated, and obtained entropy value is saved on the coordinates (x, y) of a resulting image. Here, in a noted pixel (x, y), x is ranged from 0 to (an image width minus a window width) and y is ranged from 0 to (an image height minus a window height). By using an entropy filter, a smooth portion in image quality is extracted from an original image. A preferred example of the smooth portion is a cell nucleus.
Claims7
A method of detecting a smooth area or coarse area in image quality from an original image by processing the original image using an entropy filter, said entropy filter comprising a computer and software embodied on a computer-readable memory for implementation by the computer, comprising the steps of:
- providing the original image to the computer, each pixel of said image having an intensity and a location within the image (x, y);
- calculating entropy of a small section of said image whose location in the image is derived from the location of a noted pixel (x, y) located within the small section to obtain an entropy value for said pixel (x, y) and obtaining the entropy value for said each pixel of said image, wherein entropy is calculated based on the following equation (1):
- (
- entropy
-
-
- value
- )
- =
- -
- ∑
- l
- =
- l
- min
- l
- max
-
- P
-
- (
- l
- )
-
- log
-
-
- P
-
- (
- l
- )
- (
- 1
- )
- where P(l) is a normalized intensity histogram obtained by dividing the frequency of each intensity level by the total number of frequencies, renewing the intensity of said each pixel of said image by the entropy value, and binarizing said renewed intensity of said each pixel using a threshold to obtain a binarized image including one or more detected smooth areas or coarse areas.
A method of detecting a nucleus area from a cell image by a computer, comprising:
- providing the cell image to the computer, each pixel of said cell image having an intensity and a location within the image (x, y);
- calculating entropy of a small section of said image whose location in the image is derived from the location of a noted pixel (x, y) located within the small section to obtain an entropy value for said pixel (x, y) and obtaining the entropy value for said each pixel of said image, wherein entropy is calculated based on the following equation (1):
- (
- entropy
-
-
- value
- )
- =
- -
- ∑
- l
- =
- l
- min
- l
- max
-
- P
-
- (
- l
- )
-
- log
-
-
- P
-
- (
- l
- )
- (
- 1
- )
- where P(l) is a normalized intensity histogram obtained by dividing the frequency of each intensity level by the total number of frequencies, renewing the intensity of said each pixel of said image by the entropy value, and binarizing said renewed intensity of each pixel using a threshold to obtain a binarized image including one or more detected nucleus areas.
Dependencies are inferred from the claim text.