User Guide

User Guide Neat Image
Copyright © 1999-2003 by ABSoft. All rights reserved.
5 20-Nov-03
2. Key concepts
2.1. What it can do – functionality of Neat Image
Neat Image is a digital image filter. Its main function is to reduce noise in digital images.
Neat Image can work with many imaging devices – digital cameras, scanners, etc. The program can be
adjusted to a particular device by means of a device noise profile, which contains data describing the
noise characteristics of the device working in some mode.
A device noise profile is built through analysis of featureless image areas. The key idea behind analysis
is to let the human user specify image areas that contain no visible (or important) details. The program
alone cannot distinguish noise (or some other unwanted details, like periodic interference) from
important image details while a human user can easily select an area that contains no important details.
By analyzing specified area, the program can build a device noise profile which describes this particular
noise. With the profile, Neat Image can efficiently reduce noise in the image. Also, the software can
automatically select the device noise profile best matching a particular image.
The Neat Image noise filter processes images in three frequency ranges. This makes possible reducing
noise in one frequency range even if details are present in other ranges.
Neat Image can use several working color spaces (RGB, YCrCb JPEG or YCrCb Symmetric) to process
images. Choosing suitable working color space increases the filter efficiency as well. For example,
color spaces of the YCrCb family separate the brightness and color (luminance and chrominance) image
components so it is easier to deal only with the brightness component, which often contains the major
part of visible noise.
The characteristics of noise often strongly depend on the local brightness of an image area. Neat Image
can measure these dependencies and take them into account within the filtration procedure resulting in
increased accuracy of the noise reduction.
In addition to noise reduction, Neat Image has the sharpening filter, which only sharpens important
image details without increasing the level of noise. This filter uses the same device noise profile as the
noise filter, so applying two filters together saves time and produces better overall results.
2.2. When it works – types of input images
Neat Image is designed to reduce noise in images produced by digital cameras and scanners, and can
also be used to process images from other sources. The input image should satisfy the following
requirements:
• Noise must be uniformly distributed throughout the image, i.e., there should be no strong
surges of noise intensity in some areas of the image or significant changes of noise characteristics
across the image.
Neat Image works fine, for example, on images with high ISO noise. However, ‘hot’ or ‘dead’
pixels (produced by single ‘broken’ image sensor elements) do not satisfy the uniformity
condition and, therefore, are not efficiently removed by Neat Image
1
.
Another frequent source of noise is JPEG compression. The JPEG noise is approximately uniform
when high quality compression (low compression rate) is used. However, low quality compression
makes noise non-uniform. Therefore, we recommend using the highest quality of compression
whenever possible. Try to avoid visible artifacts ('squares' or ‘blocks’ introduced by JPEG
compression) in input images beginning from the early stages of image processing!
• Noise should be concentrated in high and medium frequencies. This condition is usually met
by images produced by modern digital cameras. This condition may not be completely satisfied if
you use the strong (e.g., x2-x3 and more) digital zoom features of digital cameras.
• If you are going to use the noise profile equalizer then the primary sensors of an image
acquisition device should be RGB (as compared with image sensors with subtractive primary
colors). This requirement is not strict and may be dropped completely in the future.
1
Hot pixel removal is in our development plans.