|
|
|
Image Processing
|
The statistical correlation with invariant moments for
Geometric correction improvement
Sompong WISETPHANICHKIJ, Kobchai DEJHAN, Somkid HANPIPATPONGSA,
Fusak CHEEVASUVIT, Somsak MITATHA, Chanchai PIENVIJARNPONG*,
Chatcharin SOONYEEKAN**
Faculty of Engineering and Reserch Center for Communication and Information Technology,
King Mongkut's Institute of Technology Ladkrabang, Ladkrabang,
Bangkok 10520, Thailand.
Tel : 66-2-3269967, 66-2-3269081, Fax : 66-2-3269086
-535254,
E-MAIL:kobchai@telelan.telecom.kmitl.ac.th
* National Research Council of Thailand,
MOS-I Receiving Ground Station, Ladkrabang, Bangkok 10520, Thailand.
** Aeronautical Radio of Thailand,
102 Ngamduplee, Tungmahamek, Bangkok 10120, Thailand.
Keywords: Statistical Correlation, Geometric Correction
Abstract The systematic and non-systematic (or random) distortions are major causes of geometric distortion on remote sensing images. The reliableness of all applications usage lay down on geometric correction methods. The precision method is images registration with ground control points, but the assigned GCPs position on the corrupted images with interference noise and burry cloud coverage is difficult and bore some task. Therefore, this paper presents a method to assign automatic GCPs by using image registration method based on sequential similarity detection algorithm (SSDAs) technique, called statistic correlation measurement. The subimage around GCPs are selected and applied with invariant moments method as prototype image to assign a precise positions of GCPs are without translation, rotation and scale change problems. The above sub-images will be applied with whitening filter, that are the results of statistic correlation measurement analysis under the assumption of Markov-process before image registration, geometric transformation and resampling to carry out a precision geometric correction.
1. Introduction
The systematic and non-systematic (or random) distortions are geometric distortion that appear on satellite image and necessary to correct before using. With the ground control points(GCPs), the precision of geometric correction will be done. The corrupted-image with cloud coverage and noise make mapping between GCPs and coincident areas on the image diffculty. To automatic register the sub-image around GCP from the improved image acting as prototype image to assign the positions of GCPs for the improving satellite image, the correlation will also be used in order to obtain the precision and accuracy. The geometric correction process geometrically converts the image coordinates from (x,y) to ) , (y x , where corrected coordinates without geometrical distortion are expressed by ) , (y x and input image corrdinates by (x,y). For Precision geometric correction could be preformed the image to image registration that one image is refered to be the reference image and assumed to be of good quality, i.e., no clound to present, contrast is good and geometrical distortion are negligible. The search (input) image, on the other hand, is relatively unknown quality. Some fog/clound cover may be present along with geometrical distortion. The important issue to achieve such process is the way to select the ground control point (GCP), which should be evident on both images or topographic map, time invariance and spread through out the image. For the sake of two images of the same scene cannot be meaningfully compared. Principal among these are cross-correlation, normalized cross-correlation as shown in eq.1 and minimum distance cirteria (Webber, 1973).
Equation -1
where (j,k) are indices in a JxK point window area W, that is located within an MxN point search area S.
Figure 1. Relationship between search and window areas.
With translation, rotation and scale chane problems, the window sub-images around GCPs are selected from corrected image and appiled with non-variant moments technique to carry out the precise positions of GCPs in searching area in un-corrected image as mention in section 3. Then, the window sub-image will be applied with whitening fillter based on statistical correlatrion technique.
|
|
|
|
|
|
|
