IMAGE PROCESSING SYSTEM AND METHOD

• US 20150235100A1
• Filed: 08/07/2013
• Published: 08/20/2015
• Est. Priority Date: 07/11/2013
• Status: Active Grant

• Current Assignee: Foundation of Soongsil University-Industry Cooperation
• Original Assignee: Foundation of Soongsil University-Industry Cooperation
• Inventors: Hong, Min-Cheol, Nguyen, Tuan-Anh, Kim, Beomsu
• Granted Patent: US 9,607,238 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06F 18/24 Classification techniques
• CPC Class Codes: G06T 2207/20012 Locally adaptive
• CPC Class Codes: G06T 5/00 Image enhancement or restoration
• CPC Class Codes: G06T 5/70 Denoising; Smoothing
• CPC Class Codes: G06T 7/41 based on statistical description of texture

Abstract

An image processing system and method is provided. The image processing system includes extracting a first local statistical characteristics from an observed image, generating a low-pass filter according to the first local statistical characteristics to generate a predicted noise image, and extracting second local statistical characteristics from the predicted noise image, extracting a third local statistical characteristics from a predicted original image using the first local statistical characteristics of the observed image and the second local statistical characteristics of the predicted noise image, and setting a flag level indicating a noise level for each pixel of the observed image according to the third local statistical characteristics of the predicted original image to detect noise, and setting a filter coefficient according to the flag level of the detected noise to remove the noise, and restoring the observed image.

Claims

1. An image processing system comprising:

a noise detection unit comprising,

an observed image local statistical extraction unit that extracts a first local statistical characteristics from an observed image,

a low-pass filter design unit that designs a low-pass Gaussian filter according to the first local statistical characteristics from the observed image,

a predicted noise image generation unit that generates a predicted noise image using the low-pass Gaussian filter,

a predicted noise image local statistical extraction unit that extracts a second local statistical characteristics from the predicted noise image,

a predicted original image local statistical extraction unit that extracts a third local statistical characteristics from a predicted original image using the first local statistical characteristics of the observed image and the second local statistical characteristics of the predicted noise image, and

a noise classification unit that sets a flag level indicating a noise level for each pixel of the observed image according to the third local statistical characteristics of the predicted original image; and

a noise removal unit that sets a filter size according to the flag level and that removes noise, when the flag level for the each pixel of the observed image is determined at the noise classification unit.

2. The image processing system according to claim 1, wherein the predicted noise image generation unit generates the predicted noise image using the low-pass Gaussian filter by calculating the observed image and a filter coefficient designed by the low-pass filter design unit.

3. The image processing system according to claim 1, wherein the predicted original image local statistical extraction unit extracts the third local statistical characteristics from the predicted original image using the first local statistical characteristics of the observed image and the second local statistical characteristics of the predicted noise image by calculating:

a local average of the predicted original image, wherein the local average of the predicted original image is calculated by a local average of the observed image and a local average of the predicted noise image, and

a local standard deviation of the original image, wherein the local standard deviation of the original image is calculated by a local standard deviation of the observed image and a local standard deviation of the predicted noise image.

4. The image processing system according to claim 1, wherein the noise classification unit sets the flag level for the each pixel of the observed image according to a difference between a local average of the predicted original image and a pixel value of the observed image.

5. The image processing system according to claim 4, wherein, when the difference between the local average of the predicted original image and the pixel value of the observed image is less than a predetermined first threshold value, the flag level is set as a low level.

6. The image processing system according to claim 5, wherein, when the difference between the local average of the predicted original image and the pixel value of the observed image is greater than the predetermined first threshold value and smaller than a predetermined second threshold value, the flag level is set as a middle level.

7. The image processing system according to claim 6, wherein, when the difference between the local average of the predicted original image and the pixel value of the observed image is greater than the predetermined second threshold value, the flag level is set as a high level.

8. The image processing system according to claim 1, wherein the noise removal unit includes a noise removal filter generation unit that changes a parameter and a filter window size according to the flag level which is classified by the noise classification unit and indicates predicted noise characteristics.

9. The image processing system according to claim 8, wherein the noise removal unit further includes an image restoration unit that restores an image using a filter coefficient generated by the noise removal filter generation unit.

10. An image processing method comprising:

extracting a first local statistical characteristics from an observed image;

generating a low-pass filter according to the first local statistical characteristics to generate a predicted noise image;

extracting a second local statistical characteristics from the predicted noise image;

extracting a third local statistical characteristics from a predicted original image using the first local statistical characteristics of the observed image and the second local statistical characteristics of the predicted noise image;

setting a flag level indicating a noise level for each pixel of the observed image according to the third local statistical characteristics of the predicted original image so as to detect noise;

setting a filter coefficient according to the flag level of the detected noise to remove the noise; and

restoring the observed image.

11. The image processing method according to claim 10, wherein the setting of the flag level further comprises:

setting the flag level as a low level when a difference between a local average of the predicted original image and a pixel value of the observed image is less than a predetermined first threshold value,

setting the flag level as a middle level when the difference between the local average of the predicted original image and the pixel value of the observed image is greater than the predetermined first threshold value and less than a predetermined second threshold value, and

setting the flag level as a high level when the difference between the local average of the predicted original image and the pixel value of the observed image is greater than the predetermined second threshold value.