WEBVTT

NOTE Paragraph

1
00:00:04.100 --> 00:00:08.900
There are many key factors that go into producing the highest quality images.

2
00:00:10.800 --> 00:00:18.400
This video series will demonstrate the appearance of each image quality factor
and explain how they can be measured using Imatest.

3
00:00:19.000 --> 00:00:28.000
Noise is an undesirable random spatial variation,
visible as grain in film, or pixel level fluctuation in digital images.

4
00:00:29.300 --> 00:00:34.000
Imatest measurements help you understand the noise in your imaging system.

5
00:00:35.200 --> 00:00:41.800
They also empower engineers to develop appropriate techniques and algorithms to reduce noise.

6
00:00:45.000 --> 00:00:49.800
ISO sensitivity plays a leading role in introducing noise in a photograph.

7
00:00:50.500 --> 00:01:00.600
When photographing low light environments with a mobile device or a camera
on an automatic setting, you may notice a significant increase in the amount of noise.

8
00:01:01.200 --> 00:01:06.800
This is due to the sensor compensating for less light with increased electronic gain.

9
00:01:07.300 --> 00:01:17.900
Reducing ISO will decrease noise, but in order to keep the same exposure level,
there must be a corresponding increase in aperture size or exposure time.

10
00:01:21.300 --> 00:01:31.300
Temporal noise is  intrinsic in the imaging process,and originates from
the particle nature of light, imperfect electronics, and the thermal energy of heat.

11
00:01:34.600 --> 00:01:42.400
Non-temporal noise comes from hot or dead pixels,
or from silicon imperfections in the production of the image sensor.

12
00:01:43.400 --> 00:01:49.500
Fixed pattern noise is essentially constant.
Once it is known, if can be fully compensated for.

13
00:01:52.600 --> 00:01:58.300
Grayscale stepcharts are used to measure pixel variation across different luminance levels.

14
00:01:58.800 --> 00:02:05.900
Brightness is perceived on a logarithmic scale,
so slight changes of light become more visible in darker areas.

15
00:02:06.700 --> 00:02:11.800
This means that a constant amount of noise is more apparent at lower signal levels.

16
00:02:12.300 --> 00:02:19.800
A common and useful way to describe the impact of noise at each luminance level
is the signal to noise ratio.

17
00:02:20.400 --> 00:02:30.200
SNR is typically plotted with respect to signal level
and is often represented as a ratio or in decibels (dB).

18
00:02:30.800 --> 00:02:41.900
Noise can be measured with several Imatest modules and test charts including:
eSFR ISO, Multicharts, Multitest, Stepchart and Colorcheck.

19
00:02:45.100 --> 00:02:50.300
Measuring RAW image noise lets you learn about the performance of the image sensor itself.

20
00:02:50.900 --> 00:02:49.500
MTF50 is a commonly used metric which summarizes an MTF curve.

21
00:02:50.000 --> 00:03:01.000
A remarkable property of RAW images is that the noise in a patch of
identically-illuminated pixels is a function of pixel level and does not depend on color.

22
00:03:01.600 --> 00:03:11.100
This property is no longer necessarily true once an image has been processed,
making noise measurements extremely dependent on signal processing.

23
00:03:11.600 --> 00:03:17.100
If possible, it is useful to measure noise from both RAW and processed images.

24
00:03:20.300 --> 00:03:26.100
Measuring noise in a processed image
lets you measure how well your noise reduction has performed.

25
00:03:26.600 --> 00:03:37.200
Image signal processors (ISP’s) will reduce noisein areas that are darker
or have lower contrast, and apply sharpening to areas with high contrast.

26
00:03:40.300 --> 00:03:49.200
The effects of noise reduction can be analyzed using log-f contrast,
or with random pattern charts like Dead Leaves and Spilled Coins.

27
00:03:50.000 --> 00:03:56.100
Fine texture details in an image are typically indistinguishable from noise to most algorithms.

28
00:03:56.700 --> 00:04:05.600
Texture analysis allows you to measure how well your noise reduction
removes undesirable noise while maintaining desired fine detail.

29
00:04:08.500 --> 00:04:15.500
Depending on how a human observer views an images,
certain frequencies of noise will be less visible to the observer. 

30
00:04:16.000 --> 00:04:26.300
Visual noise metrics, developed as part of CPIQ and ISO standards,
report the effective amount of noise according to the human visual system’s perception.

31
00:04:28.300 --> 00:04:37.700
To learn more about the appearance and testing process
for other key image quality factors, please visit imatest.com/IQFactors.

32
00:04:38.400 --> 00:04:44.300
For a free 30-day trial, please visit Imatest.com/free-trial.