全文小结：
- 1、The Study of Image Restoration Techniques for Aerial Radiometric Data
- 2、如何区分空间不变模糊和空间变化模糊？
- 3、blured 在图像中是什么意思....
- 4、matlab图像锐化
The Study of Image Restoration Techniques for Aerial Radiometric Data
Zhang Yu jun
（Research Institute of Areo-Geophysics,Center of Acro-Geophysics and Remote Sensing, Minjstry of Geology and Mineral Resouces,Beijing 100083,China）
This paperpresents a very specific method for restoration of images of airborne radiometric data.High-lights of this paper indude the advancement of the principles and theory, the establishment of the processing flow-diagram, the formulation of the means for reestablishment of the gridded data file, the evaluation of the restoration results and the errors involved.
Key words: Aeroradiometric data, Atmospheric background, Image processing, lmage restoration techniques.
1 INTRODUCTION
Since the adoption of prismatic NaI（T1）crystals in the early 70's, the sensitivity and effectiveness of airbome radiometric surveying have increased significantly.The demands on aerial radiometrics by geologists and geophysicists have also changed.
During the past 20 years, the atmospheric radon background（or atmospheric background for short）could never he accurately corrected.This has been the main problem affecting the successful application of the airborne-radiomctric method.This problem manifests itself through the presence of stripes on maps and this seriously affects the usefulness of these maps.The reason can be summarized[1]as follows: the radioac-tivity measured in the air comes from the earth and also from the airframe itself,from cosmtc radiation and from atmospheric radon and its products.The latter is referred to as the atmospheric background and is influ-enced by changes in the seasonal climate, in the wind-force and wind direction, in the temperature, in the time of the day, etc.The interference of atmospheric background has different levels from flight to flight.The uranium channel suffers the most, the potassium channel is next.Although the thorium channel and the total count channel are least effected, their errors should not be neglected（see Fig.3 and Fig.4 in color plate 7）.As a result of this type of interference, the geologic informationfrom the earth is often masked.Fig.3 a（in color.plate 7）shows the restored image composed of the three elements: K（red）, Th（green）, U（blue）in the Hamitudun survey region.Fig.3 b（colored plate 7）is the composite image of the raw data.Fig.4a（colored plate 7）shows the data taken in the morning and evening for use in the correction of the flight data.Fig.4b（colored plate 7）is the raw data image of thetotal count channel.The existence of the banding noise can be compared to a striped colored veil hanging against the image of useful in-formation.With the presence of this severe noisy disturbance, it is almost impossible to generate any accept able contour maps using the raw radiomctric data from this survey region.
Noise on radiometric maps is a“world-wide”problem[2].The removal of the atmospheric background has been discussed by a number of authors.For regions having many bodies of water, Darnley and Grasty[1]recommended background corrections based on counts collected during frequentflights over large lakes.Foote[3]used an upward-looking detector shielded from ground radiation and multiple flights to estimate the atmospheric radiation in the uranium channel.Later in 1986, Grasty[4]suggested using the average of the normal section of flight line, instead of the background, when no lakes are present in the survey region.
The method introduced here is entirely different from the methods used by others.This method is an image restoration technique for aerial radiometric data.According to digital image processing terminology, image restoration is commonly defined as the reconstruction or estimation of an image field to correct image degradation and to approximate as closely as possible an ideal degradation.freeimagefield by using priori knowledge of degradation.The procedure of restoration is to model the image degradation effects and then perform operations to“undo”the model, producing an image which has been restored to a certain degree.
Dr.Cannon[5]developed a pattern removal technique or image restoration technique with the capability of removing a fingerprint from a regular pattern（fabric）, improved defocusing, capable of overcoming severe detector-to-detector noise on Landsat MSS image, clarification of motion blurred image, and so on.A similar study has been reported by Srinivasan[6].Zhang Yu-jun etal.[7]studied the image restoration problem in photos of deep-sea manganese nodules from the angle of light degradation due to a nonhomogeneous light-source.
The image restoration technique for aerial radiometrie data is a new application of the digital image restoration technique in geology.The degradation of the image of aerial radiometric data is specific and different from the above mentioned instances.This method has been tested and verified in preliminary research for 6 survey regions.
2 PRINCIPLE AND THEORY OF IMAGE RESTORATION TECHNIQUE FOR AERIAL RADIOMETRIC DATA
A degraded image G（x,y）is obtained by aerial radiometric survey.It can be regarded as the sum of the degradation-free ideal image F（x,y）and an interfering imageη（x,y），The degradation process can be simplified and is shown in Fig.1.
Fig.1 Diagram for degradation of imago of aerial radiometric data.
The priori knowledge of degradation for aerial radiometric data image can be obtained by the analysis of an aerial radiometric survey and the raw data image.During the survey,the information originating from the geologic bodies is independent of time,but the interfering signal is essentially time dependent.On the image the interfering signal can be represented as a function of（x, y）asfollows:
but
so
张玉君地质勘查新方法研究论文集
张玉君地质勘查新方法研究论文集
张玉君地质勘查新方法研究论文集
The change ofηcan be divided into two parts: the stepped change between flights and the gradual change within a flight, see Fig.4（in color plate 7）.The interfering signal can be considered as a constant on each flight line T, If x（the column on an image）expoesses the direction perpendicular to the flight line, the functionη（x,y）may be simplified into（x），thus
张玉君地质勘查新方法研究论文集
The purpose of restoration of the aerial radiometric image is simply to find an approximateη（x）and to approxjmately obtain the F（x, y）.In this connection, convolution can be conducted several times along the flight line direction for the raw data image using a long narrow window with several lines and a single column,leading to:
张玉君地质勘查新方法研究论文集
W is the plate of convolution, and it is a weighting matrix.Convolution is a type of linear operation, the operatorHis space-invariant.Since the operator is linear, the operation is additive.Thus, the response of the sum of two inputs equals the sam of the two respones
张玉君地质勘查新方法研究论文集
Since it is assumed that thefunctionηonly correlates with x, and the convolution window is a singlecolumn one,
张玉君地质勘查新方法研究论文集
The characteristics of function HF（x, y）will now be examined.Since a smooth average has been performed severaltimes along the y direction, the local anomalies are almost“drowned out”by the appar ent regional features which manifest themselves as a slow change along the flight line.If the local anomaly is expressed byf（x,y）and the apparent regional field byL（x,y），we have
张玉君地质勘查新方法研究论文集
after thefollowing process
张玉君地质勘查新方法研究论文集
It can be seen from equation（9）that the restored imagef（x,y）which is a result of the raw data image by the substraction of the noise image, closely approximates the ideal image from the point of view of the local anomalies.The error depends on the amplitude of change of the substracted“apparent regional background”along the flight line direction.
3 PROCEDURE FOR RESTORATION HANDLING OF AERIAL RADIOMETRIC DATA IMAGE
The study of the image restoration techniquefor aerial radiometric data is based on the theory of multivariate statistical analysis.It is accomplished by means of image processing.It demonstrates image processing as a directly visual and fast procedure.Its flow-chart is illustrated in Fig.2.
The method assumes that the noise background of aerial radiometric data is non-variable or only linear-ly variable, By a smooth averaging conducted several times along the flight line direction, the local anoma-lies will gradually disappear as they are“drowned out”by the noise background.The resultant noise image is linearly correlated with the noise background and subject to some edge compensation.After the substraction of noise, we conducted a median filtering and a Wallis transformation（space variant contrast stretch）.This finally led to the restoration effect.This restoration process is shown in the left-hand half of Fig.2.
The right-hand half of Fig.2 shows the reconstruction process of the gridded data file, which is indispensable in routine application.After classification and region separation we can get the mean-vectors for all classes before and after restoration.By a least-squares fitting we can obtain the element concentration values or the count-rates for the restored image.A gridded data file for contouring on the main-frame computer can finally be produced by the inverse transformation.
Fig.2 Procedure for restoration of handljng of aerial radiometric data image.
In this investigation we have also tried to get the noise levels by averaging all datafor each flight line, but the results are not as ideal as those gained by the above described method.
4 RESULTS AND ERROR EVALUATION
4.1 ResuIts of the I mage Restoration for AeriaI Radiometric Data
（1）Improved direct visual effect on maps
The image restoration process bears a striking analogy to the drawing aside of a striped colored veil to disclose the original clean features of the gamma-spectrum data that are hidden behind and obscured by the veil（see Fig.3 in color plate 7）.The sawtooth-shaped events occurring on the boundaries of some geologic bodies as aresult of inaccurate positioning of flight lines,bothforward and backward, are also much ameliorated（see Fig.5 in color plate 7）.Fig.5（colored plate 7）displays the comparison imagefor the total count channel:Fig.5a shows the raw data, Fig.5b the noise image,Fig.5c the noise-removed image and Fig.5d illustrates the restored image.
（2）The upgraded contour maps after data restoration
The contour maps of K,Th and U channels from the Hamitudun survey could not be drawn at first on the main-flame computer because of the severe banding interferences in the raw data.Only stacked profiles were supplied for these elements.Although the contour map for the total count channel was drawn,the banding effect was still distinctly visible.
By applying the technique of image restoration and by reconstructing the gridded data files and feeding them back to the main-frame computer,good quality contour maps were produced for TC, K,Th and U.Fig.6（color plate 7）.shows the contour map for K channel after restoration.Its anomalies have a good correlation with the geologic bodies on the geological map.There is also a good agreement between the radioactivities of the anomalies and the lithology.All this proves the effectiveness of this technique.The quality of the contour maps has been significantly upgraded by the process.The accuracy of these maps is further confirmed by the classification map using restored data（see Fig.7 in color plate 7）.In the classification image of Fig.7 there are 9 classes:1.ultrabasic,2.basic,3.granite,4.diorite,5.metamorphic,6.migmatite,7.Quaternary sediments,8.Tertiary and Quaternary sediments and 9.Tertiary sediments.
（3）The enrichment of useful information
Multivariate statistica1analysis has been used in this studyfor the quantitative eva luation of the effectof image restorationfor aerial radiometric data.It is possible to evaluate this effect by the variation of useful information in some images.For this purpose it is necessary to calculate the mean variation,which is the average of the total variation for a single pixel.The symbols C,C´and G" express the mean variations for useful information in the primary image,for the interference information in the primary image andfor the useful information in the restored jmage,respectively.
In the statistics,G´（x,y）is approximately taken asη（x），while[G（x,y）－G´（x,y）]showsF（x,y）approximately,and P（x,y）represents thefinal restored image.It is assumed that there are no errors.
张玉君地质勘查新方法研究论文集
The letters with a bar above them indicate the mean values.M,Nstandfor the numbers of lines and columns in an image.
Table 1 illustrates the statistical results for quantitative evaluation by using the above listedformula for the aerial radiometric data images of the Hamitudun survey region.
Table 1 demonstrates the remarkable increase of useful informationfor all the K,Th,U and TC restored images.Speaking about this survey region,the quality of primary images for TC and K channels is better than those for Th and U channels.
Table1 Statistical results for quantitative evaluation of the aerial radiometric data images of the Hamitudun survey region
4.2 The Evaluation of Accuracy and Error for the Restored I mages
The main error in a restored image comes from the“apparent regional background”L（x, y）formed by smooth averaging done several times.The following evaluationfor accuracy was found by the statistics of the profide data on the interfering images.
K±0.16%（absolute concentration）Th±2.1 ppm
U±0.15ppm TC±869.6counts
5 CONCLUSIONS
1.The method described here was suggested in China and overseas as a new specific technique of image restoration for aerial radiometric data.Its reliability and practicability were tested by data in several survey regions.
2.This technique can basically remove the banding noise on,radiometric data maps caused by the changing of atmospheric background and by unstable thresholds.It can basically restore the ideal image of the aerial radiometric data.It also provides preparation for further image processing（such as:enhancement, gradientation,classification,logic operation,etc.）.Therefore,it can be used as a quick method for data pretreatment.
3.This method can improve the sawtooth-shaped noise on the image at the boundaries of some geologic bodies as a result of inaccurate positioning.
4.The mean variation of useful information is suggested in this study as a measure for quantitative evaluation of restoration effects for the aerial radiometric data image.Also discussed in this study is the absolute error or accuracy of the method with which image restoration is used as a pre-treatment process.
ACKNOWLEDGEMENTS
Many people contributed to the success of this work.In particular,I would like to mention Lin Zhen-min for his valuable discussions,Shi Jian-wen for his taking part in the repeated tests,Zhang Zhi-min and Xie Xin who developed the programs for data transformation and least-squaresfitting, Shui En-hai who collected the correction data in the testing region,Lu Lin-sheng,who helped with English,and Li Wei-hua, who typed the text on a word processor.I am very grateful to all of them.
REFERENCES
[1]Grasty,R.L,Gamma ray spectrometric methods in uranium exploration-Theory and operational procedures,Geophysics and Geochemistry in the Search for Metallic Ores,GSC Ottawa,147-162,1977.
[2]Green,A.A.,Leveling airborne gamma-radiation data using between-channel correlation information,Geophysics,52,1557-1562,1987.
[3]Foote,R.S.,Improvement in airborne gamma-radiation data analysis by removal of environ-mental and pedologic radiation changes,in Sympos.on the Use of Nudear Techniques in Prospecting and Development of Mineral Resources: Internat.Atomic Energy Agency Mtg., Buenos Aires, 187-196,1968.
[4]Grasty,R.L.,Automated system for computing on-line atmospheric backgrounds,GSC paper, 1-52,1987.
[5]Cannon,M.,Lehar,A.and Preston,F.,Background pattern removal by power spectral filtering,Applied Optics,22,777-779,1983.
[6]Srinivasan,R.,Software image restoration techniques,Digital Design,16,4,29-34,1986.[7]Zhang Yu-jun and Shi Jian-wen,A study of image reconstruction and image processing techniques for photos of deep-sea polymetallic nodules,Geophysical and Geochemical Exploration（inChinese）13,435-441,1989.
原载《Chinese Journal of Geophysics》，1990,Vol.33,No.3.
如何区分空间不变模糊和空间变化模糊？
大概说的应该是这个意思。空间不变 就是输出仅和输入有关和位置无关。空间变化输出则和位置相关
blured 在图像中是什么意思....
少了个r？blurred 模念配糊
blurred image 模糊图像仔缺指
motion-blurred image运动扮高模糊图像
matlab图像锐化
%%%目测大岩含你的是拉普拉斯算子 下面的程序也是一滚笑样的 你试试 图片改成你的图片
A=imread('lena.bmp');
figure(1);
subplot(1,2,1);
imshow(A);
title('原图');
I=double(A);
h=[-1 -1 -1;-1 9 -1;-1 -1 -1];
J=conv2(I,h,'same');
K=uint8(J);
subplot(1,2,2);
imshow(K);
title('使用拉普拉斯算子锐化枣陪处理后的图');
