Method of reducing noise in image and image processing apparatus using the same
Provided are methods and apparatuses for reducing noise in an image, in which a kernel that gives a weight to a pixel value with respect to determination of a similarity between patches is adaptively changed according to a reference patch, a non-transitory computer-readable storage medium for recording the same, and an image processing apparatus using the same, in which noise in an image is removed according to a noise removal algorithm based on non-local means.
1. A method of reducing noise in an image, the method comprising:
setting a plurality of pixels in the image as a reference patch, and setting target patches in the image corresponding to the reference patch, the reference patch having a center pixel;
determining a similarity within the reference patch between the center pixel in the reference patch and nearby pixels in the reference patch;
adaptively setting a kernel based on the similarity within the reference patch;
determining a similarity between the reference patch and the target patches using the kernel; and
removing noise from the center pixel in the reference patch based on pixel values of center pixels in the target patches by applying, a weight to the center pixels of the target patches based on the similarity between the reference patch and the target patches.
2. The method of claim 1 , wherein the kernel is adaptively changed according to a distribution of pixel values of the reference patch.
3. The method of claim 1 , wherein the kernel gives a relatively high weight to a pixel value of the center pixel in the reference patch, if a similarity between the center pixel in the reference patch and the nearby pixels is high.
4. The method of claim 1 , wherein the kernel is generated based on difference values between the center pixel in the reference patch and the nearby pixels in the reference patch that are close to the center pixel and a maximum one of the difference values and a minimum one of the difference values.
5. The method of claim 4 , wherein the kernel is generated by normalizing the difference values between the center pixel in the reference patch and the nearby pixels using the maximum value and the minimum value.
6. The method of claim 4 , wherein the kernel is generated based on location differences from the center pixel in the reference patch to the nearby pixels.
7. The method of claim 1 , wherein the determining the similarity between the reference patch and the target patches further comprises:
calculating a difference between pixels in the reference patch and pixels in the target patch that corresponds to the pixels in the reference patch.
8. The method of claim 1 , wherein the setting sets the target patches in images adjacent in an image sequence to the image.
9. A non-transitory computer-readable storage medium having stored thereon a computer program, which when executed by a computer, configures the computer to perform the method of claim 1 .
10. An image processing apparatus configured to reduce noise in an image, the image processing apparatus comprising:
a processor configured to,
set a plurality of pixels in the image as a reference patch, and to set target patches in the image corresponding to the reference patch, the reference patch having a center pixel,
determine a similarity within the reference patch between the center pixel in the reference patch and nearby pixels in the reference patch,
adaptively set a kernel based on the similarity within the reference patch,
determine a similarity between the reference patch and the target patches using the kernel, and
remove noise from the center pixel in the reference patch based on pixel values of center pixels in the target patches by applying a weight to the center pixels of the target patches based on the similarity between the reference patch and the target patches.
11. The image processing apparatus of claim 10 , wherein the processor is configured to adaptively change the kernel according to a distribution of pixel values of the reference patch.
12. The image processing apparatus of claim 10 , wherein the kernel gives a relatively high weight to a pixel value of the center pixel in the reference patch if a similarity between the center pixel in the reference patch and the nearby pixels is high.
13. The image processing apparatus of claim 10 , wherein the processor is configured to generate the kernel based on difference values between the center pixel in the reference patch and the nearby pixels in the reference patch that are close to the center pixel and a maximum one of the difference values and a minimum one of the difference values.
14. The image processing apparatus of claim 13 , the processor is configured to generate the kernel by normalizing the difference values between the center pixel in the reference patch and the nearby pixels using the maximum value and the minimum value.
15. The image processing apparatus of claim 13 , wherein the processor is configured to generate the kernel based on location differences from the center pixel in the reference patch to the nearby pixels.
16. The image processing apparatus of claim 10 , wherein the processor is configured to further calculate a difference between pixels in the reference patch and pixels in the target patch that corresponds to the pixels in the reference patch.
17. The image processing apparatus of claim 10 , wherein the processor is configured to further set the target patches in images adjacent in an image sequence to the image.
18. The image processing apparatus of claim 10 , wherein the image is a depth image representing a distance between an image photographing apparatus and a subject.