THE USE OF WAVELETS FOR NOISE DETECTION IN THE IMAGES 
TAKEN BY THE ANALOG AND DIGITAL PHOTOGRAMMETRIC CAMERAS 
K. Pyka 2 3, *, J. Siedlik b 
a Department of Geoinformation, Photogrammetry and Remote Sensing of Environment AGH University of Science 
and Technology Krakow- krisfoto@agh.edu.pl 
b Business Group MGGP AERO Tamow -jsiedlik@mggpaero.com 
Commission I, WG 1/1 
KEY WORDS: Noise, Wavelets Transform, Decomposition,, Detail Component, Kurtosis, Variance 
ABSTRACT: 
In the paper the use of wavelet transformation for valorization of random noise content in photogrammetric images is proposed. 
There were two wavelets indicators studied. The first indicators based on the analysis of the wavelet detail coefficients distribution 
shape. As the results prove, noise conclusions based on the shape are not always objective. As the second noise indicator the 
analysis of changing of relative variance during decomposition was researched. Based on the studies, it has been proven that the 
analysis of the equation of preservation of image relative variance is a good indication of the noise level. The low noise level is 
proven by a stable increase of the details variance along with the level of decomposition. In case of fine-grained image texture, such 
increase is undisturbed. For the research a set of aerial images taken by two photogrammetric cameras, analogue LMK and digital 
DMC was compared. In all examined cases the better parameters of the noise evaluation were obtained for the digital camera. The 
researches confirmed the possibility to define the noise content indicators based on the analysis of the wavelet detail coefficients. 
1. INTRODUCTION 
The radiometric quality is often neglected when looking for the 
reasons of the unsatisfactory quality of the automatic images 
analysis. Meanwhile, the image noise can substantially reduce 
the efficiency of image processing, especially in the case of 
images with low contrast and containing many fragments with 
fine-grained texture, with which we deal often in 
photogrammetry and remote sensing. 
We have witnessed the process of replacing the analogue 
photogrammetric cameras with digital ones. Direct digital 
image acquisition reduces the number of stages in which noise 
can occur, but does not liquidate the problem of its occurrence. 
Noise should be regarded as an immanent feature of the 
photogrammetric images, similarly to, for example, lens 
distortion. 
For some years the discrete wavelet transformation has been 
used in the image processing. The wavelet transformation is 
regarded as the most effective method of lossy compression of 
multitonal images and is more and more often used in the 
photogrammetric working stations. In this paper the using of 
wavelets for evaluation of the image random noise is proposed. 
2. THE NEED FOR IMAGE NOISE INDICATORS 
Noise is any random or deterministic disturbance of luminance 
of a hypothetical image that would come into existence in the 
ideal conditions (Morain, 2004). Image noise together with 
radiometric resolution, contrast, tonal matching are elements 
shaping radiometric quality. Noise arises in the different stages 
of the image acquisition: during the image forming, sampling, 
encoding, compression, transmission and during image 
processing. 
Random noise is present practically in any image, but is not 
always noticed. In the analogue images, the source of random 
noise is the granular structure of photographic emulsion. Noise 
in the digital images is caused by instability of detectors, 
including - to some extent - detector’s own noise. 
Random noise, because of its unpredictable character, cannot be 
removed completely from the image. One can only smooth over 
the effects of its occurrence. We face a dilemma: Is it better to 
reduce the noise level at the expense of the edges sharpness or 
the other way round? The photogrammetric and remote sensing 
multitonal landscape images, which are taken from large 
distances predominate, have frequently low local contrast and 
small signal-to-noise ratio. This is why there is a need to look 
for the indicators of noise content. 
3. THE USEFULLNESS OF WAVELET FOR NOISE 
DETECTION - THEORETICAL STUDY 
3.1 Basic features of wavelet transformation 
Wavelet transformation demonstrates some features shared with 
Fourier transformation. The Fourier transformation converts the 
signal from spatial domain into frequency domain. The wavelet 
transform is a frequency-spatial representation, i.e. it is possible 
Corresponding author.