Noise filtering method and apparatus considering noise variance and motion detection

Patent metadata

Field	Value
Publication number	US-9070185-B2
Application number	US-201113286409-A
Country	US
Kind code	B2
Filing date	Nov 1, 2011
Priority date	Dec 28, 2010
Publication date	Jun 30, 2015
Grant date	Jun 30, 2015

Abstract

Provided herein are an apparatus and a method of performing image noise filtering with respect to an image obtained from an image sensor, the method including: applying a noise deviation with respect to a temporal difference between a previous pixel value and a current pixel value when obtaining a temporal weight value to be used in temporal filtering. The noise deviation is obtained according to an intensity level of the current pixel value. Temporal filtering and spatial filtering may be performed, and blending filtering may be selectively performed by blending an output value of the temporal filtering and an output value of the spatial filtering.

First claim

What is claimed is:

1. An image noise filtering method comprising:
obtaining a temporal weight based on an adaptive noise deviation value obtained with respect to a temporal difference between a current pixel and a previous pixel;
refining the obtained temporal weight based on a peripheral pixel of the current pixel and other channel pixels; and
performing temporal filtering on the current pixel based on the refined temporal weight,
wherein an output value of the temporal filtering is blended with an output value of spatial filtering, to obtain a result of spatio-temporal filtering as a sum of the output value of the temporal filtering and a weighted difference between the output value of the spatial filtering and the output value of the temporal filtering, and
the output value of the spatial filtering is obtained by blending a value of the current pixel and a mean value of peripheral pixels in a spatial domain using the temporal weight.

2. An image noise filtering method comprising:
obtaining a noise deviation according to an intensity level of a current pixel value;
obtaining a temporal weight by applying the noise deviation with respect to a temporal difference between a previous pixel value and the current pixel value;
performing temporal filtering by blending the current pixel value and the previous pixel value using the temporal weight;
performing spatial filtering by blending the current pixel value and a mean value of peripheral pixels in a spatial domain using the temporal weight;
performing blending filtering by blending an output value of the temporal filtering and an output value of the spatial filtering; and
selecting one of an output of the blending filtering and an output of the spatial filtering in response to a selection control signal,
wherein the performing blending filtering comprises:
calculating a result of blending as a sum of the output value of the temporal filtering and a weighted difference between the output values of the spatial filtering and the temporal filtering.

3. The image noise filtering method of claim 2 , further comprising:
selecting one of the output of the blending filtering, the output of the spatial filtering, and the current pixel value in response to the selection control signal.

4. The image noise filtering method of claim 2 , further comprising:
refining the temporal weight according to a peripheral pixel of the current pixel and other channel pixels, before the temporal filtering is performed.

5. The image noise filtering method of claim 2 , further comprising:
encoding the temporal weight when the temporal weight is stored to process a pixel on a next frame.

6. The image noise filtering method of claim 5 , further comprising:
decoding the encoded temporal weight.

7. The image noise filtering method of claim 2 , wherein the noise deviation varies according to a characteristic of an image sensor.

8. The image noise filtering method of claim 2 , wherein the image noise filtering method is performed by one of a digital camera and a handheld communication terminal.

9. An image noise filtering device comprising:
an intensity level calculating part which calculates an intensity level of a current pixel value;
a noise standard deviation calculating part which calculates a noise deviation according to the intensity level;
a temporal weight value generating part which generates a temporal weight by applying the noise deviation with respect to a temporal difference between a previous pixel value and the current pixel value;
a temporal blending part which blends the current pixel value and the previous pixel value using the temporal weight;
a spatial filtering part which blends a mean value of peripheral pixels and the current pixel value in a spatial domain using the temporal weight; and
a spatio-temporal blending part configured to blend an output value of the temporal blending part and an output value of the spatial filtering part and calculate a result as a sum of the output value of the temporal filtering and a weighted difference between the output values of the spatial filtering and the temporal filtering.

10. The image noise filtering device of claim 9 , wherein the temporal blending part is configured to calculate a result of blending of the current pixel value and the previous pixel as a sum of the current pixel value and a weighted difference between the previous pixel value and the current pixel value.

11. The image noise filtering method of claim 1 , wherein the obtaining the temporal weight comprises:
determining noise deviation curves which define relationships between pixel intensity values and a standard noise deviation for different lighting conditions;
storing data corresponding to the noise deviation curves which differ from one another in correspondence with the different lighting conditions; and
applying a different value of the standard noise deviation, with respect to an intensity value of the temporal difference between the current pixel and the previous pixel, by using the stored data of the noise deviation curves.

Assignees

• Lee Sanghoon
• Park Sungcheol
• Samsung Electronics Co Ltd

Inventors

• Lee Sanghoon
• Park Sungcheol

Classifications

• G06T5/002 Primary

  Physics · mapped topic

• G06T2207/20182

  Noise reduction or smoothing in the temporal domain; Spatio-temporal filtering · CPC title

• G06T5/70 Primary

  Denoising; Smoothing · CPC title

• H04N5/21

  Circuitry for suppressing or minimising disturbance, e.g. moiré or halo · CPC title