IMAGE QUALITY RELATED PROCESSING AND APPLICATIONS BASED ON
RETINEX WAVELET THEORY
Qingwu HU *
School of Remote Sensing and Information Engineering, Wuhan University, Wuhan, P.R. China, 430079
KEY WORDS: Remote Sensing, Image, Quality, Multi-resolution, Wavelet Transforms, Mosaic
ABSTRACT:
In this paper, a new image processing model based on Retinex technique and multi-resolution wavelet analysis together to simulate
the imaging course and focus mechanism of HSV is proposed. At first, the image presentation based on the Retinex model of the
lightness and colour perception of human vision, which is related to the neuropsychological functions of individual neurons in the
primate retina, lateral geniculation nucleus, and cerebral cortex, is designed to decompose image into constant part and variance part
with ill-problem. Secondly, the multi-resolution wavelet transformation is proposed as a multiple resolution observation and
analyses of image to express the image with a changeable course in wavelet domain which is sensitive to the eyc's psychology
character of the human vision view effect and its adjustable focus performance. The corresponding imaging dodging enhancement
and restoration algorithms in the Retinex wavelet (RW) domain are designed to improve image quality with different image defects.
The image processing principle based on the Retinex wavelet analysis theory is exposed together with image processing applications
of image mosaic of photogrammetry workstation. The large area mosaic experiment with different luminance and uneven tone image
shows it can remove mosaic scar and obtain a seamless mosaic orthodox-image with smooth tone and colour transitions after the
Retinex wavelet restoration processing.
1. INTRODUCTION
The unavoidable image quality problems become more and
more outstanding affection to all kinds of image applications,
which greatly affect the development and application of image
understanding and image vision, such as automatic mapping,
object automatic identification and artificial intelligence. Image
qualities related processing is based on the common imaging
model as equation 1(F.M.J. Starck, 1998, Rank, 1999).
M-1N-1
g(x,v)= > > f(m,n)h(x ^ m,y - n) * n(x, y) (1)
m=0 n=0
Where g(x, y) is quality-descending image, f(x,y) is the
real scene image, n(x, y) is the random noise and A(x, y) is
imaging model. As we know, some part area in a degrade image
or some objects in a degrade image can obtain good imaging
and view effect even the surround area or other objects is
imaged worse quality. In the other side, a degrade image with
bad geometric shape characteristic maybe has well radiant
feature. The image processing should be selective while the
image processing based on the above traditional model is done
the total image without differing image part with different
operator. In fact, imaging course or imaging system is simulated
to human vision system (HSV) and many researches on image
processing technique combine the principle and characteristic of
HSV (A. Blake, 1987). E. H. Land, 1971, 1977 proposed
Retinex technique for imaging analysis and high quality image
presentation, which have obtained marketable applications for
improving classification accuracy of remote images, medical
image enhancement to detect cancer (Z. Rahman, 2001). While
Retinex technique is only to way of image presentation methods
using constant decomposition, the image analysis should be
done in the multi-resolution level, which is simulated to HSV.
Wavelet analysis provides a multi-resolution image analysis
tool for its best changeable focus performance, which is
simulated to focus function of human optic nerve. Many
classical methods have been proposed to solve such problems,
including several multi-scale and wavelet based approaches (N.
Bose, 1998, M. Elad, 1999, B. Tom, 2000). Retinex theory
proposals an image representation accordant to imaging course
while it has not the function of multiple resolution image
analysis (D.J. Jobson, 1997, Z. Rahman, 1997. 1998, B.
Thompson, 2000). Image analysis based on wavelet should be
done with a suitable image representation to obtain best image
quality.
In this paper, a new image processing model based on Retinex
technique and multi-resolution wavelet analysis together to
simulate the imaging course and focus mechanism of HSV is
proposed. At first, the image presentation based on the Retinex
model of the lightness and colour perception of human vision,
which is related to the neuropsychological functions of
individual neurons in the primate retina, lateral geniculation
nucleus, and cerebral cortex, is designed to decompose image
into constant part and variance part with ill-problem. Secondly,
the multi-resolution wavelet transformation is proposed as a
multiple resolution observation and analyses of image to
express the image with a changeable course in wavelet domain
which is sensitive to the eye's psychology character of the
human vision view effect and its adjustable focus performance.
The corresponding imaging dodging enhancement and
restoration algorithms in the Retinex wavelet (RW) domain are
designed to improve image quality with different image defects.
The image processing principle based on the Retinex wavelet
analysis theory is exposed together with image processing
applications of image mosaic of photogrammetry workstation.
The large area mosaic experiment with different luminance and
uneven tone image shows it can remove mosaic scar and obtain
a seamless mosaic orthodox-image with smooth tone and colour
transitions after the Retinex wavelet restoration processing.
2. METHODOLOGY
Current image quality related processing algorithms and
application cannot radically solve the uniform of image quality
improvement, image quality assessment and different imaging
sensors. The overall technique proposed in this paper is to
obtain best imaging representation model simulated to imaging
course and imaging analysis method of HSV, which it is called
* Corresponding author: Tel.:0086-27-87664633; Fax: :0086-27-87664633; Email: qwhu@263.net
Intei
J
e
Ny
+ OS
rx
