1) The document discusses using an autocorrelation function (ACF) filter on burst image sequences to reduce noise in CMOS image sensors and achieve high quality imaging. 2) It explains that the ACF calculates correlation values based on pixel values sampled over time to distinguish random noise from true signals. Noise pixels will have lower ACF values while true signals have higher values near 1. 3) The algorithm judges each pixel, applying a leveling filter only to pixels below thresholds for value and ACF. This reduces random noise without impacting bright pixels and resolution. Results show noise reduction while maintaining detail.

1. Noise reduction in CMOS image sensorsNoise reduction in CMOS image sensors
for high quality imaging: Thefor high quality imaging: The
autocorrelation function filter on burstautocorrelation function filter on burst
image sequencesimage sequences
Kazuhiro Hoshino1, Frank Nielsen2,3, Toshihiro Nishimura4
1 Image Sensor Business Group, Sony Corporation,
4-14-1 Asahi-chou, Atsugi-shi, Kanagawa, Japan
Kazuhiro.Hoshino@jp.sony.com,
2 Sony Computer Science Laboratories, Inc.
3-14-13 Higashi Gotanda, Shinagawa-ku, Tokyo, Japan
Frank.Nielsen@acm.org
3 Ecole Polytechnique, LIX F-91128 Palaiseau Cedex, France
4 Graduate School of Information, Production and Systems, Waseda University
2-7 Hibikino, Wakamatsu, Kitakyushu, Fukuoka, Japan
toshi-hiro@waseda.jp

2. n
VDecoder
TG
(Condenser, CDS, Decoder)
Programmable
Gain Amp
Dark noise （ C ）
Dark shot noise （ W ）
Photon shot noise （ W ）
Condensing Gm （ C ）
Amp noise （ W ）
Offset noise （ C ）
Reset noise （ W ）
1/f noise(W ）
Image pixel
Amp noise （ W ）
Offset noise （ C ）
Condensing
Gm （ C ）
1/f noise （ W ）
Analog circuit
Noise in image sensor
RST
SEL
Tx
Col. Bus
N
P
N+
CMOS image sensor W white noise, C colored noise.

3. Principle of an ACFPrinciple of an ACF
 The data is collected at the same interval time.The data is collected at the same interval time.
 Autocorrelation value is calculated according to the following equation.Autocorrelation value is calculated according to the following equation.
R is ACF value.R is ACF value.
N is the number of data,N is the number of data,
t is time.t is time.
x is pixel value,x is pixel value,
and τ is shifted time.and τ is shifted time.
∑
−−
=
+=
1
0
)()(
1
)(
τ
ττ
N
t
txtx
N
R

4. 1D simulation of ACF1D simulation of ACF
(A) cosine wave (B) white noise wave
Original
wave
ACF
value
Make
base wave
Make
base wave
Make
noise wave
Make
noise wave
Sampling
In same
interval time
Sampling
In same
interval time＋
＋
(A) cosine wave
(B) white noise wave
white noise wave
Calculation
ACF value
Calculation
ACF value
Block diagram of 1-D ACF methodBlock diagram of 1-D ACF method

5. Expansion ACF method to 2-D model
Time(a.u)
H direction
Ｉｍａｇｅ
Time
H
V
R is ACF value.R is ACF value.
N is the number of data which were sampled in time axis,N is the number of data which were sampled in time axis,
t is time.t is time.
x is pixel value,x is pixel value,
and τ is shifted time.and τ is shifted time.
∑
−−
=
+=
1
0
)()(
1
)(
τ
ττ
N
t
txtx
N
R

6. ACF value as a function of pixel intensityACF value as a function of pixel intensity
Flame Number
Pixel- Ａ
Pixel-B
Bright pixel A (180 in 256 scale) and dark pixel B (8 in 256 scale)
AutoCorrelationValue
(A)
(B)

7. The algorithm of noise judging and filtering processThe algorithm of noise judging and filtering process
by a time domain ACF methodby a time domain ACF method
Image data (BMP,RAW)
Pixel value extraction
Calculation of ACF
Leveling filter processing
END
Pixel value decision
Pixel value i<10
ACF value r<0.8
I< Total pixel number
No
No
No

8. Result of image processingResult of image processing
Original image Processing image
・ Reduction of random noise is possible per pixel.
・ Since filter processing is not performed in a bright pixel, resolution does not
deteriorate.

9. Image data (BMP,RAW)
Pixel value extraction
Calculation of ACF
Leveling filter processing
END
Pixel value decision
Pixel value i<10
ACF value r<0.8
I< Total pixel number
No
No
No
The algorithm and the example of processing of a time
domain ACF method

10. Original Ith= 100
Rth=0.985
Ith= 100
Rth=0.985
Ith= 100
Rth=0.995
Ith= 100
Rth=1.000
Image processing result as a function of threshold
value both pixel value and ACF value