INTERNATIONAL JOURNAL FOR TRENDS IN ENGINEERING & TECHNOLOGY VOLUME 5 ISSUE 1 – MAY 2015 - ISSN: 2349 - 9303
Retrieving Informations from Satellite Images by Detecting and Removing Shadow T.Dhanya Priya1
K.Vidhya2
Sri Ramakrishna Engineering College, dhanyapriya93@gmail.com
Sri Ramakrishna Engineering College, vidhyasrec@gmail.com
Abstract— In accordance with the characteristics of remote sensing images, we put forward a color intensity method of shadow detection and removal. Some approaches for shadow detection and removal use particular color and spectral properties of shadows. In this method, the input satellite image color plane is calculated and the values of RGB are separated. Then the chromaticity is calculated to determine the average value of the segmented region. The Color Intensity algorithm is adopted to remove the shadow and retrieve the corresponding information.
Keywords— chromaticity, color intense algorithm, region of interest, remote sensing, shadow detection, shadow removal. —————————— —————————— for detection and de-shadowing, respectively. Research on I. INTRODUCTION shadow correction is still an important topic, particularly for urban regions and mountains. Existing shadow detection High resolution satellite image has become an increasingly methods can be roughly categorized into two groups : modelimportant source of information, mainly for applications where based methods and shadow-feature-based methods. The first the need for details is essential as is urban environment. In the group uses prior information such as scene, moving targets, last ten years, with the availability of high-spatial-resolution and camera altitude to construct shadow models. This group of satellites such as IKONOS, QuickBird, GeoEye, and Resource methods is often used in some specific scene conditions such as 3 for the observation of Earth and the rapid development of aerial image analysis and video monitoring. The second group some aerial platforms such as airships and unmanned aerial of methods identifies shadow areas with information such as vehicles, there has been an increasing need to analyze highgray scale, brightness, saturation, and texture. These resolution images for different applications. The Shadow is one approaches for shadow detection and removal use particular of the major problems in remotely sensed imaging which color and spectral properties in shadows. In order to accurately hampers the accuracy of information extraction and change identify a shadow, the threshold value is obtained from the detection. Shadows are caused by the interaction between light estimated grayscale value of the shadow areas. However, and objects, such as buildings, trees and bridges, etc. Shadows information such as scene and camera altitude is not usually can enhance the reality of images, and can often confound readily available. Consequently, most shadow detection algorithms designed to solve other vision tasks such as image algorithms are based on shadow features. For example, the segmentation or the locating and tracking objects in a scene. shadow region appears as a low grayscale value in the image, On the other hand, the poor visibilities of features in shadow and the threshold is chosen between two peaks in the grayscale regions severely degrade the interpretability of the images. histogram of the image data to separate the shadow from the Although shadows can be regarded as a type of useful nonshadow region. In a related study, images are converted information of the 3-D reconstruction, building position into different invariant color spaces (HSV, HCV, YIQ, and recognition, and height estimation, they can also interfere with YCbCr) to obtain shadows. Based on that work, a successive the processing and application of high-resolution remote thresholding scheme was proposed to detect shadows. The sensing images. For example, shadows may cause incorrect method used by Makarau et al accurately detected shadows results during change detection. Consequently, the detection with the blackbody radiation model. Recently, a hierarchical and removal of shadows play an important role in applications supervised classification scheme was used to detect shadows. of high-resolution remote sensing images such as object A variety of image enhancement methods have been classification, object recognition, change detection, and image proposed for shadow removal, such as histogram matching, fusion. In the field of remote sensing, only few works on gamma correction, linear correlation correction(LCC), and shadow detection have been carried out restoration of the color invariance model. In addition, a paired-region-based approach is employed to detect and remove the shadows in a single image by calculating the mainly concerning building detection. In recent years, difference between the shadow and nonshadow regions of thresholding and recovering techniques have become important the same type. Aside from the aforementioned methods,
55
INTERNATIONAL JOURNAL FOR TRENDS IN ENGINEERING & TECHNOLOGY VOLUME 5 ISSUE 1 – MAY 2015 - ISSN: 2349 - 9303 shadows can be retrieved using malicious data. For example, shadow pixels can be identified from the region of interest in an image and from another image obtained at a different time. Then, nonshadow pixels of the corresponding region are used to replace the shadow pixels. This latter approach is useful in low-resolution images.
III.SHADOW DETECTION AND REMOVAL Shadow detection in image still remains as a challenging problem. Separation of shadow from object area is a difficult problem for a single images. A. Methods based on type of shadow E.Salvator, A. Cavall, and T. Ebrahimi proposed an approach to detect and classify shadow for images. They exploit invariant color features to classify cast and self shadows. In the first level, they utilize edge detection followed by a morphological operation to extract object and cast shadow regions. A dark region extraction process is then applied to identify shadow candidates in the segmented region. In the second level, they obtained an edge map which does not contain the edges corresponding to shadow boundaries. The obtained edge map is used, together with the dark region map, to distinguish between self and cast shadows. This can be used to detect shadows in images. But some constraints are uniform colored object and non structured surfaces. Rosin assumed that shadow is a region with reduced contrast and shadow region is detected using a region growing algorithm. But the problem is that a region growing algorithm cannot perform accurately in the penumbra part of the shadow.
II.OVERVIEW OF SHADOW AND ITS TYPES To detect shadows, we must consider the appearance of the local and surrounding regions. Shadowed regions tend to be dark, with little texture, but some nonshadowed regions may have similar characteristics. Surrounding regions that correspond to the same material can provide much stronger evidence. There are two types of shadows: the self-shadow and the cast shadow. A self-shadow is the shadow on a subject on the side that is not directly facing the light source. A cast shadow is the shadow of a subject falling on the surface of another subject because the former subject has blocked the light source. A cast shadow consists of two parts: the umbra and the penumbra. The umbra is created because the direct light has been completely blocked, while the penumbra is created by something partly blocking the direct light. In this paper, we mainly focus on the shadows in the cast shadow area of the remote sensing images.
B. Based on color spaces
The causes of shadow in remote sensed imaging can be grouped into three categories: 1) Shadow by urban materials such as building and trees. This is a special problem in high spatial resolution imaging. 2) Shadow by mountain (topographic shadow). This can be a major difficulty in medium spatial resolution imaging and high spatial resolution imaging as well. 3) Cloud shadows. This problem can occur in high, medium, and coarse spatial resolution imaging. It is noteworthy that cloud shadow and topographic shadow are not spectrally distinguishable in optical imaging. However, their differences can be determined by geographic position. The direction of shadow occurrence relative to the cloud is the same for the entire scene as it depends on the sun illumination angle, a constant for any given scene.
56
There are different types of color spaces like RGB, YCbCr, LAB etc. Corina, Peter Jozesef, Zoltan and Laszlo presented an algorithm to automatically detect shadows and remove them. Initially they represent the picture in the YCbCr color space. They focused on the Y channel and computed its histogram. Then computing the full average of the image at Y channel , and performed sliding window iteration through the image. In order to decide which pixels belong the shadow, they used methods. The pixels that have the intensities lower than 60% of the full average are considered as shadow. The nonshadow point average is computed for the sliding window. The pixel that have intensified lower than 70% of the window’s average is considered as the part of the shadow. First one for correction on Cb and Cr channels and second for the correction on the Y channel. The shadow detection has good results at homogeneous regions, but for more textured regions it could result in false detections. K.Emily Esther Rani and G.Jemilda introduce a preprocessing step using the RGB color model to identify the presence of shadows in an image. Once the presence of shadow is confirmed, then the image is converted into a shadow invariant image. To detect the shadow boundaries, TAM(Tricolor Attenuation Model) based shadow detection algorithm is introduced. A preprocessing step to segment the original image in to sub regions with similar color is also introduced. The algorithm describes the attenuation relationship between shadow and its non-shadow regions. This method can automatically detect and extract the shadows from a single still images, complex outdoor scenes and this method do not need any user intervention and prior knowledge. If a pixel gets a sample that is likely to be a shadow, then not only the GMM of the pixel is updated, but the GMM of the neighbor pixels is also updated. Markov random fields are employed to represent the dependencies among neighbor pixels. For global level information, statistical feature is exploited for the whole scene over several consecutive frames.
INTERNATIONAL JOURNAL FOR TRENDS IN ENGINEERING & TECHNOLOGY VOLUME 5 ISSUE 1 – MAY 2015 - ISSN: 2349 - 9303 IV.PROPOSED METHOD Shadow detection can be achieved through different methods depending upon the kinds of shadows. Here, we has propose a method to detect the shadows in the satellite images. The method involves 4 major steps: Step Step Step Step
1: 2: 3: 4:
Formation of 2D Chromaticity Shadow Detection Boundary Formation Shadow Removal
A. Formation of 2D Chromaticity The input image is first checked whether it contains RGB color or not. If any noise is present, it is removed by using Gaussian filter. The RGB values are put forth as a matrix with RGB array values. The Chromaticity function (XY function) is implemented to obtain a 2D matrix. It is mainly used to segregate the data. This array function consists both the normal and the logarithmic values of X and Y.
B. Shadow Dectection In shadow detection step, the input vector value is taken into consideration. The entire image value is traced both rowwise and column-wise respectively. The traced value is rounded to the nearest values in order to obtain the accurate normalized and log normalized values of the entire array. Then the entropy function is incorporated to obtain the average value of the matrix. The rechroma function is used to re-segregate the obtained values.
C.Boundary Extraction Boundary Extraction is done to segment the shadow region from the non-shadow region. The Contour based masking technique is used to mask the shadowed region. The obtained boundary is extracted and shown by using red lines. The thickness of the red line is already defined by the values. Finally Smoothing filter is used to smooth the edges of the shadowed region.
D.Shadow Removal Removal is the last and most important step. The average value of the non-shadow region is calculated and compared with the average values of the detected shadow region and the obtained value is appended in the shadowed region. Region of Interest algorithm is implemented for the shadow removal technique.
57
INTERNATIONAL JOURNAL FOR TRENDS IN ENGINEERING & TECHNOLOGY VOLUME 5 ISSUE 1 – MAY 2015 - ISSN: 2349 - 9303 Flow chart of the proposed method.
V.EXPERIMENTAL RESULTS: The proposed shadow detection and removal algorithms were implemented in MATLAB version R2013a (8.1.0.604) under the Microsoft Windows 7 environment. We selected the satellite images of various environments to test the algorithms described in the previous sections. Five groups of results are demonstrated below. Each group is displayed by two rows.
Fig 1.a
Fig 1.b
Fig 1.c
Fig 1.d
Fig 3.a
Fig 3.b
Fig 3.c
Fig 3.d
Fig 3 Satellite Image of RKV Stadium. Fig 1 Satellite Image of Earth
Fig 2.a
Fig 4.a
Fig 4.b
Fig 4.c
Fig 4.d
Fig 2.b
Fig 4 Satellite Image of Urban area.
Fig 2.c
Fig 2.d
Fig 2 Satellite Image of Himalayas.
58
INTERNATIONAL JOURNAL FOR TRENDS IN ENGINEERING & TECHNOLOGY VOLUME 5 ISSUE 1 – MAY 2015 - ISSN: 2349 - 9303 Output Image Fig 1.d Fig 2.d Fig 3.d Fig 4.d Fig 5.d
SD value of R 44.0452 16.0068 29.7918 08.5146 24.7013
SD value of G 37.2351 16.9680 30.7960 08.0251 26.0762
SD value of B 35.9006 17.8890 38.5565 07.3764 26.7667
The table 1.2 shows the standard deviation values of the of the corresponding output images. Fig 5.a
Fig 5.b
Fig 5.c
Fig 5.d Fig 6.2 Graphical representation of output SD values.
Fig 5 Satellite Image of Urban area.
Table 2
In each group the fig a represents the input image, fig b represents the binary representation of the image, fig c represents the result of boundary Extraction and fig d represents the retrieved output image. The comparisons of the above group of results are shown in the table. Table 1.1 Input Image Fig 1.a Fig 2.a Fig 3.a Fig 4.a Fig 5.a
SD value of R 88.4835 20.3289 30.6117 08.5146 38.2690
SD value of G 86.5167 18.6670 30.4270 08.0251 38.8353
SD value of B 85.9585 17.9399 36.2183 07.3764 37.7231
Image
PSNR value of R
PSNR value of G
Fig Fig Fig Fig Fig
13.323769 19.977032 22.093144 21.578862 20.855740
13.143278 19.487905 22.643215 21.563110 21.292173
1 2 3 4 5
PSNR value of B 13.244790 19.884944 22.338559 21.989300 21.624830
The Table 2 shows the PSNR values of the images used in the experiment. As shown in the above figures, chromaticity level shadow detection method based on spectral features and spatial features can accurately and effectively detect shadows in highresolution remote sensing image. Fig.1 shows the satellite image of an earth. It is feasible to judge the trend and range of a shadow by collecting and analyzing the boundary lines of an image. Taking an automobile, for example in fig.4, black cars are regarded as suspected shadows after thresholding. By thresholding, the shadow of the white truck was judged to be a suspected shadow. It can successfully, obtain the true shadow from the suspected shadow by comparing the grayscale image of the surroundings. Therefore, using spatial relative information can remove a part of false shadow, but not all of it.
The Table 1.1 shows the standard deviation values of the various input images.
VI.CONCLUSION We have put forward a systematic and effective method of shadow detection and removal for high-resolution satellite images. In order to get a shadow detection result, image segmentation considering shadows is applied first. Then, suspected shadows are detected through chromaticity of the image. The subsequent shadow detection experiments
Fig 6.1 Graphical representation of input SD values. Table 1.2
59
INTERNATIONAL JOURNAL FOR TRENDS IN ENGINEERING & TECHNOLOGY VOLUME 5 ISSUE 1 – MAY 2015 - ISSN: 2349 - 9303 compared with the traditional experiments. Meanwhile, they also show the effects of different steps with the proposed method. For shadow removal, we put forward a color intense based algorithm. The experimental results revealed the following. 1) The shadow detection method proposed in this paper can stably and accurately identify shadows. 2) Compared with histogram detection, the chromaticity detection method proposed in this paper can make full use of the color information of an image. However, it is difficult to segment the small size shadows into an independent object, which will cause errors. 3) The shadow removal method based on color intensity can effectively restore the information in a shadow area. 4) The obtained standard deviation values of input and output images are compared.
[4]
J.Y.Rau, N.Y.Chen and L.C.Chen, ―Hidden Compensation and Shadow Enhancement for true Orthophoto Generation‖, Proc. Asian Conf.Remote Sensing,2000,p.p.112-118.
[5]
J.Sun, Y.K.Du, and Y.D.Tang, ―Shadow Detection and Removal from Solo Natural Image based on Retinex Theory‖,Proc.Int.Conf.Intelligent Robotics and Applications(ICIRA 2008), Wuhan,China,Oct.1517,2008,pp.660-668. Saritha Murali and V.K.Govindan,‖Removal of shadows from a single image‖ ,In the Proceedings of First International Conference on Futuristic Trends in Computer Science and Engineering, volume – 4,pages 111-114,ICCT 2012.
[6]
Further improvements are needed in the following ways. Although image segmentation considering shadows can have better segmentation results, insufficient segmentation still exists. For example, a black car and its shadow cannot be separated. Also, parts of the shadow from low trees cannot be separated from the leaves.
[7]
Saritha Murali ,‖Shadow Detection and Removal from a single image‖ ,MTech Project Thesis, NIT Calicut, 2011’12.
[8]
Masashi Baba and Naoki Asada, 2003, ‖Shadow removal from a real picture‖ ,ACM SIGGRAPH 2003 Sketches \& Applications, pages:1-1.
[9]
G D Finlayson, S D Hordley, and M S Drew,‖Removing shadows from images‖, In proceedings of the 7th European Confernce on Computer Vision-Part IV, pages 823-836, London, UK, Springer-Verlag 2002.
[10]
Shahtahmassebi,A., Ning,Y., Ke,W., Moore.N., Zhangquan,S., ―Review of Shadow Detection and Deshadowing Methods in Remote sensing‖ ,Springer , Chin.Georgia.Sci.Vol. 23,2013.
[11]
Golchin,M., Khalid,F., Abdullah,L., Davarpanah,S., ―Shadow Detection using Color and Edge Information‖ , Journal of Computer Science , November 2013.
[12]
Saleh,M., Al-Hadi,N., ―Shadow Removal Using Segmentation Method‖ J. Of College Of Education For Women, vol.24 (1) 2013.
References [1]
Hongya Zhang, Kaimin Sun, and Wenzhuo Li, ―ObjectOriented Shadow Detection and Removal From Urban High-Resolution Remote Sensing Images‖, IEEE Transactions on Geoscience and Remote Sensing, VOL. 52, NO. 11, November 2014.
[2]
Priya Garg and Kirtika Goyal, ―Detection and Removal of Shadow using Chromaticity‖, International Journal of Computer Science and Information Technologies, Vol. 5(4),2014,5745-5747.
[3]
Corina BLAJOVICI, Peter Jozsef KISS, Zoltan BONUS, and Laszlo VARGA. ―Shadow Detection and Removal from a Single Image‖, SSIP 2011, Szeged, Hungary.
60