automatic 3d view generation from a single 2d image for both indoor and outdoor scenes

International Journal on Computational Sciences & Applications (IJCSA) Vol.3, No.4, August 2013

DOI:10.5121/ijcsa.2013.3404 37

Automatic 3D view Generation from a Single 2D Image for both Indoor and Outdoor Scenes

Geetha Kiran A1 and Murali S

2

1Malnad College of Engineering, Hassan, Karnataka, India

[email protected]

2Maharaja Institute of Technology, Mysore, Karnataka, India

[email protected]

ABSTRACT

Image based video generation paradigms have recently emerged as an interesting problem in the field of

robotics. This paper focuses on the problem of automatic video generation of both indoor and outdoor

scenes. Automatic 3D view generation of indoor scenes mainly consist of orthogonal planes and outdoor

scenes consist of vanishing point. The algorithm infers frontier information directly from the images using

a geometric context-based segmentation scheme that uses the natural scene structure. The presence of

floor is a major cue for obtaining the termination point for the video generation of the indoor scenes and

vanishing point plays an important role in case of outdoor scenes. In both the cases, we create the

navigation by cropping the image to the desired size upto the termination point. Our approach is fully

automatic, since it needs no human intervention and finds applications, mainly in assisting autonomous

cars, virtual walk through ancient time images, in architectural sites and in forensics. Qualitative and

quantitative experiments on nearly 250 images in different scenarios show that the proposed algorithms

are more efficient and accurate.

KEYWORDS

Floor segmentation,canny edge detector, hough transform, vanishing point, video generation

1. INTRODUCTION

Video generation from a single image is inherently a challenging problem. In Imaging devices,

there is a trade-off between the images (snapshots) and video because of the limitation in storage

capacity. Video clips need more storage space compared to images. This motivated to generate

the video from a single 2D image rather than storing video clips. Humans analyze variety of

single image cues and act accordingly, unlike robots. The work is an attempt to make robots

analyze similar to humans using single 2D image. The task of generating video from photographs

is receiving increased attention in many of the applications. We are addressing here the key case

where dimension of the real world object or measurement of object dimension in 2D plane is

unknown. However generating video using above methods is very difficult because of

perspective view. Alternatively, video could be generated using proper ground known i.e., floor

segmentation in case of indoor scenes. In the absence of accurate measurements, we wish to

exploit geometric characteristics (windows/doors) along with the color variations. Such

relationships are plentiful in man-made structures and often provide sufficient information to our

work. In case of Road scenes, video could be generated using proper ground known i.e.,

vanishing point. We describe a unified framework for navigation through a single 2D image in

lesser time. The input image may be easily acquired since no calibration target is needed or we

can download images from internet. The work is well-suited for navigation on Personal Digital

assistants(PDA’s) and personal computers, includes cases where buildings are destroyed and only

the archive images are available. It can also be applied in forensics and to assist autonomous cars

by generating video from a single 2D image and assessing in advance - how far there is a straight


38

road? If there is any suspected person or item in our path of journey, it could be detected prior

and necessary action can be taken. We describe a unified framework for generating video from a

single 2D image. In the next section, a review on the related works is highlighted. Section 3 gives

description of floor segmentation from single view scene constraints along with the computation

of length of the floor. Section 4 gives description of finding the vanishing point from single view

scene constraints and computing the distance from the ground truth position to the vanishing

point. This is followed by the method of 3D view generation in section 5. Finally, some of the

experimental results are presented in section 6 followed by conclusion in section 7.

2. RELATED WORK

It is observed that some methods have been developed for segmentation on a single image, few

which are directly relevant to the work are highlighted here. Erick Delage et al have used a graph

based segmentation algorithm to generate a partition of the image and assigned a unique

identifier to each partition output by the segmentation algorithm in [18]. Erick Delage et al [20]

have built a probabilistic model that incorporates a number of local image features and tries to

reason about the chroma of the floor in each column of the image. Ma Ling et al [22] has

segmented the floor region automatically by adopting clustering analysis and also have proposed

a PCA based improved version of the algorithm to remove negative effect of shadow for

segmented results. Xue – Nan Cui et al [23] have proposed detecting and segmenting the floor

by computing plane normals from motion fields in image sequences. A geometric characteristic

that objects are placed perpendicular to the ground floor can be utilized to find the floor in 2D

images. Surfaces often have fairly uniform appearances in texture and color and thus image

segmentation algorithms provide another set of useful features which can be used in many other

applications, including video generation. Some of the methods developed for detecting

vanishing point on a single image have been highlighted here. Techniques for estimating

vanishing points can be roughly divided into two categories. One requiring the

knowledge of the internal parameters of the camera and the other operates in an

uncalibrated setting. A large literature exists on automatic detection of vanishing points,

after Barnard [1] first introduced the use of the Gaussian Sphere as an accumulation

space. He suggested that the unbounded space can be mapped into the bounded surface of

the Gaussian sphere. Tuytelaars et al [2] mapped points into different bounded subspaces

according to their coordinates. Rother [3] pointed out these methods could not preserve

the original distances between lines and points. In this method, the intersections of all

pairs of non-collinear lines are considered as accumulator cells instead of a parameter

space. But these accumulator cells are difficult to index, searching for the maximal from

the accumulator cells is slow. The simple calculation of a weighted mean of pairwise

intersection is used by Caprile et al [4]. Researches [5-7] have used vanishing point as

global constraint for road. They compute the texture orientation for each pixel and select

the effective vote-points, then locate the vanishing point by using a voting procedure. Hui

Kong et al [8-11] have proposed an adaptive soft voting scheme which is based upon a

local voting region using high-confidence voters.

However, there are some redundancies during the voting process and the accuracy on

updating vanishing point. Murali S et al [12,13] have detected edges using canny edge

detector and hough transform is applied. The maximum votes of first N number of cells

in the hough space is used for computing the vanishing point. We use the similar

framework [12-13] in our work to decide the vanishing point.A very few Researchers

have proposed different methods for navigation through a Single 2D image. Shuqiang


39

jiang et al [14] have proposed a method to automatically transform static images to

dynamic video clips in mobile devices. Xian-sheng Hua et al [15] developed a system

named photo2video to convert a photographic series into a video by simulating camera

motions. The camera motion pattern (both the key-frame sequencing scheme and

trajectory/ speed control strategy) is selected for each photograph to generate a

corresponding motion photograph clip. A region based method to generate a multiview

video from a conventional 2-dimensional video using color information to segment an

image has been proposed by Yun-Ki-Baek et al [16]. Na-Eun Yang et al [17] have

proposed method to generate depth map using local depth hypothesis and grouped

regions for 2D-to-3D conversion. The various methods of converting 2D to stereoscopic

3D images involves the fundamental, underlying principle of horizontal shifting of pixels

to create a new image so that there are horizontal disparities between the original image

and the new version. The extent of horizontal shift depends on the distance of the feature

of an object to the stereoscopic camera that the pixel represents. It also depends on the

inter-lens separation to determine the new image viewpoint.

The methods proposed by the authors for floor segmentation is time consuming and have made

certain assumptions specific to the application. These artifacts are not of much importance in our

work, this made us to propose a simple method for floor segmentation in lesser time. Using the

segmented image, length of the floor could be computed by distance method. This helps in video

generation. The methods proposed by the authors for detecting vanishing points have made

certain assumptions specific to the application. These artifacts are not of much importance in our

work, this made us to propose a new method as proposed in [12,13], which decides the vanishing

point in lesser time. Using the vanishing point, the distance from the ground truth position to the

vanishing point could be computed. This helps in navigating through the single Road image.

3. FLOOR SEGMENTATION

The goal is to obtain floor segmentation of a given single 2D indoor image. The crucial part of

the work is detecting the pixels belonging to the floor. There are methods available for floor

segmentation with known camera parameters. Requirements is to segment floor without having

knowledge of camera parameters. There is possibility to find the geometric relationship, may be

using color. The primary steps involves converting the given color image to gray, further convert

the gray image to binary image by computing a global threshold. Finally, segment the floor by

applying the dilation and erosion methods.

3.1. Segmentation

The floor path [19] is the major cue to generate video from a single 2D image of indoor scenes.

To segment the floor from the remaining parts of the indoor image scenes, dilation and erosion

techniques using the structuring elements are used.

Assuming E to be a Euclidean space or an integer grid, A a binary image in E, and B a structuring

element.

The dilation of A by B is defined by:

(1)


40

The erosion of A by B is given by:

(2)

Structuring element is used for probing and expanding the shapes contained in the input image

yielding to floor segmentation (Figure 1).

(a) (b) (c) (d)

Figure 1. (a) Original Image (b) Gray Image

(c) Binary Image using Otsu’s method (d) Segmented Image

The segmented image obtained ( Figure 1(d) ) is used to find the length of the floor. The distance

between the start and end of the white pixel (row wise) from the floor segmented image is found

by using the Euclidean distance method. This length of the floor identified could directly be used

to decide the number of frames to be generated, generally 1:2 depends on the length and it can be

varied with requirements. These frames are incorporated in the video generation.

4. VANISHING POINT DETECTION

Images considered for modeling are perspective. In a perspective image, lines parallel in the

world space appear to meet at a point called Vanishing point. Vanishing points provide a strong

geometric cue for inferring information about 3 dimensional structure of a scene in almost all

kinds of man-made environment. There are methods available for detecting vanishing points

with known camera parameters and also with uncalibrated setting. The method described in this

section requires no knowledge of the camera parameters and proceeds directly from geometric

relationships. The step involves detecting edges using canny edge technique to identify the

straight lines depending upon the threshold fixed by the hough transform, compute the vanishing

point using the intersection points of the lines. The above steps have been explained in the

subsequent sections.

4.1. Line Determination

The given color image is converted to gray. Lines are edges of the objects and environment

present in an image. These lines may or may not contribute to form the actual vanishing point.

The existence of the lines are obtained by applying the canny edge detection algorithm. The

versatility of the canny algorithm is to adapt to various parameters like the sizes of the Gaussian

filter and the thresholds. This will allow it to be used in detecting edges of differing

characteristics.


41

The input image (Figure 2(a)) is converted to gray image (Figure 2(b)), the edges are detected by

applying Canny edge detection algorithm. A set of white pixels containing edges are obtained

and the rest of the contents of the image are removed. A Canny edge detected image (Figure

2(c)) contains pixels contributing to straight lines and also other miscellaneous edges.

Considering all these pixels of the edges contributing to the straight lines, Hough transformation

is applied on the input image and the result (Figure 2(d)) is obtained as desired.

(a) (b) (c) (d)

Figure 2. (a) Original Image (b) Gray Image (c) Edge detection (d) Hough transformation

As the outcome of the Hough transformation, a large number of straight lines are detected. These

straight lines depend upon the threshold fixed up for the Hough transformation. Points belonging

to the same straight line in the image plane have corresponding sinusoids which intersect in a

single point in the polar space (Figure 1(d)). The need for calculating the number of straight lines

is that there could be several straight lines in the image which intersects each other at different

points in the image plane. In such case there arises a situation that more than one peak value in

the polar space is obtained. Thus by selecting the number of peak values (in descending order of

their votes) equal to the number of straight lines ‘N’ present in the image we restrict the

unwanted lines which may not contribute to the real vanishing point. This reduces the

computational complexity of vanishing point detection to only the number of straight lines

contributing the possible vanishing point.

4.2. Intersection Point of any Two Lines

Lines drawn by Hough transformation are on edges of the object and environment in an image.

These lines may or may not contribute to form the actual vanishing point. Depending upon the

number of lines present in the image, the number of peaks in the Hough space is fixed up in a

descending order of their occurrences. Each peak in the hough space signifies the existence of a

longer edge in the image than any other points in the Hough space and hence a peak is formed.

These peaks of the voted points of the hough space are calculated to find the intersection between

two lines to calculate the vanishing point. Finding the intersection points for all combination of

lines selecting two at a time, corresponding one (x,y) pair is obtained. The number of pairs of x

and y values obtained for all combinations is given by the relation

(3)

where N is the number of peaks selected. These (x,y) pairs are the probable vanishing points. All

of them are within the vicinity of the actual vanishing point. We have taken the mean of the


42

probable vanishing points (Figure 3(a)-blue color), since they are within the vicinity of the actual

vanishing point. In our work vanishing point is used to find the distance from the ground truth

position to the detected Vanishing point position. The distance obtained is used in the next

section to facilitate the termination point for the navigation. The distance of the Road identified

could directly be used to decide the number of frames to be generated, generally 1:2 depends on

the length and it can be varied with requirements.

(a) (b)

Figure 3.(a) Lines detected (green Color),Probable Vanishing Points(blue Color)

(b) Vanishing Point (Pink Color)

5. 3D VIEW GENERATION

Automatic 3D view generation from a single image is inherently a challenging problem. The

proposed method has attempted to generate the 3D view (Algorithm) for both indoor and outdoor

scenes.

Algorithm

Step 1: Read the Input Image

Step 2: Compute the termination point using

i ) Floor segmentation for indoor scenes

ii)Vanishing point for outdoor scenes

Step 3: Generate the frames based on the predefined rectangle

Step 4: Navigate through the single 2D Image upto the termination point

5.1 Indoor Scenes

The information obtained in the floor segmentation is used to generate the 3D view. The input for

the video generation are: single 2D image, computed termination point based on the distance

calculated using floor segmentation, the size of the rectangle based on which cropping takes

place. The input image is considered as the first frame and the image is cropped based on the

size of the predefined rectangle. The rectangle has to be clearly defined as it is the frame further

used for 3D view generation. The floor segmentation plays a vital role in detecting the

termination point to generate the frames. Then the cropped image is resized to the original image


43

and stored in an array of images. An appropriate set of key-frames (Figure 4) are determined for

each image based on the distance computed by using floor segmentation.

(a) (b) (c)

(d) (e) (f)

Figure 4. (a) 20th Frame (b) 40

th Frame (c) 60

th Frame(d) 80

th Frame (e) 100

th Frame

(f) 120th Frame

5.2 Outdoor Scenes (Road Scenes)

The information obtained from section 4 is used to navigate through a single Road image. The

input for the navigation are - single 2D image, computed termination point based on the distance

from the ground truth position to the detected vanishing point. Based on this strategy, the frames

for navigation are generated by cropping the image based on the size of the image up to the

computed distance. The input image is considered as the first frame and the image is cropped

based on the size of the predefined rectangle. Then the cropped image is resized to the original

image and stored in an array of images. An appropriate set of key-frames (Figure 5) are

determined for each image based on the distance computed by using vanishing point.

(a) (b)


44

(c) (d)

(e) (f)

Figure 5. (a) 10

th Frame (b) 20

th Frame (c) 40

th Frame(d) 70

th Frame (e) 100

th Frame

(f) 180th Frame

6. EXPERIMENTAL RESULTS

The algorithm is applied to a test set of 250 images obtained from different buildings, all of them

are fairly different in interior decoration themes from each other. Since the indoor images

contained a diverse range of orthogonal geometries (wall posters, doors, windows, boxes,

cabinets etc.), we have observed that the results presented are indicative of the algorithm

performance on images of new buildings (interior) and scenes. We also have evaluated the

algorithm( Figure 6(a)) by manually detecting the floor path of a set of images and compared it

with the floor path generated by our method and the overall accuracy obtained from the result is

91.46%. In case of outdoor scenes obtained from different real-road images in different scenarios

that mainly consist of single vanishing point, we have observed that the results presented are

indicative of the algorithm performance. The images used in the experimentation are downloaded

from internet and few of them are self captured. The steps involves detecting edges using canny

edge technique, to identify the straight lines, compute the vanishing point using the intersection

points of the lines. All of them are within the vicinity of the actual vanishing point. Based on the

ground truth position, we compute the distance from the ground truth position to the computed

vanishing point. We also have evaluated the algorithm (figure 6(b)) by manually detecting the

distance from the ground truth value to the vanishing point and compared it with the distance

generated by our method and the overall accuracy obtained from the result is 97%.


45

(a) (b)

Figure 6. (a) Comparison of the length of the floor computed manually with our method

(b) Comparison of vanishing point accuracy computed manually with our method

The first, intermediate and final frame generated by the methods for both Indoor and Road

Scenes (Figure 7) after deducing the termination point are clearly viewed. The frames give the

finer details in the intermediate and final frames that could be used in various applications

including virtual walk through ancient time images, in forensics, in architectural sites and in

automated vehicle.

7. CONCLUSION

An algorithm for automatic video generation from a single 2D image is proposed and

experimented for both indoor and outdoor images. This paper provides a solution to transform

static single 2D image into video clips. It not only helps the users to enjoy the important details of

the image but also provides a vivid viewing manner. The experimental results show that the

algorithm is performing well on a number of indoor and outdoor scenes. The work is

experimented on nearly 250 images in difficult scenarios. Further work can be extended to

produce videos including side view, working at planar level. This requires maintenance of

perspective view of the scene. Further work may be extended to include investigating on more

reliable Region Of Interest (ROI) detection techniques. Even finer details can be obtained from

the key frames used in video generation. The work is done in view of assisting the automated

vehicle and robots at low cost.


46

REFERENCES [1] BARNARD S T, (1983) “INTERPRETING PERSPECTIVE IMAGES”; ARTIFICIAL INTELLIGENCE, 21,

PP.435-462.

[2] TUYTELAARS T, VAN GOOL L, PROESMANS M, MOONS T, (1998) “THE CASCADED HOUGH

TRANSFORM AS AN AID IN AERIAL IMAGE INTERPRETATION”; IN: PROC. INTERNATIONAL CONFERENCE ON

COMPUTER VISION, PP.67-72.

[3] ROTHER C,(2002) “A NEW APPROACH FOR VANISHING POINT DETECTION IN ARCHITECTURAL

ENVIRONMENTS”; IMAGE AND VISION COMPUTING 20, PP.647-655.

[4] B.CAPRILE AND V TORRE, (1990) “USING VANISHING POINTS FOR CAMERA CALIBRATION”;

INTERNATIONAL JOURNAL OF COMPUTER VISION, 4, PP.127-139

[5] NICHOLAS SIMOND, PATRICK RIVES, (2003) “HOMOGRAPHY FROM A VANISHING POINT IN URBAN

SCENCES”; IN PROCEEDINGS OF IEEE/RSJ INTERNATIONAL CONFERENCE ON INTELLIGENT ROBOTS AND

SYSTEMS, PP.1005-1010.

[6] CHRISTOPHER RASMUSSEN, (2004) “GROUPING DOMINANT ORIENTATIONS FOR ILL-STRUCTURED

ROAD”; IN PROCEEDINGS OF IEEE INTERNATIONAL CONFERENCE ON COMPUTER VISION AND PATTERN

RECOGNITION, PP.470-477.

[7] CHRISTOPHER RASMUSSEN, THOMMEN KORAH, (2005) “ON-VEHICLE AND AERIAL TEXTURE

ANALYSIS FOR VISION-BASED DESERT ROAD”; IN PROCEEDINGS OF INTERNATIONAL WORKSHOP ON

COMPUTER VISION AND PATTERN RECOGNITION, PP. 66-71.

[8] HUI KONG, JEAN-YUES AUDIBERT, JEAN PONCE(2009) “ VANISHING POINT DETECTION FOR ROAD

DETECTION”; IN PROCEEEDINGS OF IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION,

PP.96-103.

[9] HUI KONG, JEAN-YUES AUDIBERT(2010) “JEAN PONCE, GENERAL ROAD DETECTION FROM A

SINGLE IMAGE”; IN PROCEEDINGS OF IEEE TRANSACTIONS ON IMAGE PROCESSING, VOL.19, NO.8,PP.2211-

2220.

[10] M NIETO AND L SALSGDO,(2007) “REAL-TIME VANISHING POINT ESTIMATION IN ROAD SEQUENCES

USING ADAPTIVE STEERABLE FILTER BANKS”; ADVANCED NOTES IN COMPUTER SCIENCE.

[11] CHRISTOPHER RASMUSSEN,(2004) “ TEXTURE-BASED VANISHING POINT VOTING FOR ROAD SHAPE

ESTIMATION”; BMVC.

[12] AVINASH AND MURALI S,(2005), “ A VOTING SCHEME FOR INVERSE HOUGH TRANSFORM BASED

VANISHING POINT DETERMINATION”; IN PROCEEDINGS OF INTERNATIONAL CONFERENCE ON COGNITION

AND RECOGNITION, MYSORE, INDIA.

[13] AVINASH AND MURALI S,(2007) “ MUTIPLE VANISHING POINT DETERMINATION”; IN PROCEEDINGS

OF IEEE INTERNATIONAL CONFERENCE ON COMPUTER VISION AND INFORMATION TECHNOLOGY,

AURANGABAD, INDIA.

[14] SHUQIANG JIANG AND HUIYING LIU AND ZHAO ZHAO AND INGMING HUANG AND WEN GAO,

(2007) “GENERATING VIDEO SEQUENCE FROM PHOTO IMAGE FOR MOBILE SCREENS BY CONTENT

ANALYSIS”; ICME, PP.1475-1478.

[15] XIAN-SHENG HUA AND LIE LU AND HONG-JIANG ZHANG,(2004) “AUTOMATICALLY CONVERTING

PHOTOGRAPHIC SERIES INTO VIDEO”; 12TH ACM INTERNATIONAL CONFERENCE ON MULTIMEDIA,PP.708-

715.

[16] YUN-KI BAEK, YOUNG-HO SEO,DONG-WOOK KIM AND JI-SANG YOO,(2012) “MULTIVIEW VIDEO

GENERATION FROM 2-DIMENSIONAL VIDEO”; INTERNATIONAL JOURNAL OF INNOVATIVE COMPUTING,

INFORMATION AND CONTROL, VOL 8, NUMBER 5(A), PP. 3135-3148.

[17] NA-EUN YANG, JI WON LEE, RAE-HONG PARK,(2012) “ DEPTH MAP GENERATION FROM A SINGLE

IMAGE USING LOCAL DEPTH HYPOTHESIS”; 2012 IEEE ICCE ,PP.311-312.

[18] ERICK DELAGE HONGLAK LEE ANDREW Y. NG, (2006) ”A DYNAMIC BAYESIAN NETWORK MODEL

FOR AUTONOMOUS 3D RECONSTRUCTION FROM A SINGLE INDOOR IMAGE”; CVPR.

[19] GEETHA KIRAN,A.AND MURALI,S, (2013) “AUTOMATIC VIDEO GENERATION USING FLOOR

SEGMENTATION FROM A SINGLE 2D IMAGE” 21ST WSCG 2013 CONFERENCE ON COMPUTER GRAPHICS,

VISUALIZATION AND COMPUTER VISION, ISBN 978-80-86943-76-3.

[20] ERICK DELAGE, HONGLAK LEE, AND ANDREW Y. NG, (2005)” AUTOMATIC SINGLE-IMAGE 3D

RECONSTRUCTIONS OF INDOOR MANHATTAN WORLD SCENES”; ISRR.

[21] D. HOIEM, A. A. EFROS, AND M. HEBERT,(2005) ”GEOMETRIC CONTEXT FROM A SINGLE IMAGE”;

10TH IEEE INTERNATIONAL CONFERENCE ON COMPUTER VISION.


47

[22] MA LING , WANG JIANMING ; ZHANG BO ; WANG SHENGBEI,(2010) ”AUTOMATIC FLOOR

SEGMENTATION FOR INDOOR ROBOT NAVIGATION”; 2ND INTERNATIONAL CONFERENCE ON SIGNAL

PROCESSING SYSTEMS (ICSPS),PP. 684 – 689.

[23] XUE-NAN CUI, YOUNG-GEUN KIM, AND HAKIL KIM,(2009) ” FLOOR SEGMENTATION BY

COMPUTING PLANE NORMALS FROM IMAGE MOTION FIELDS FOR VISUAL NAVIGATION”; INTERNATIONAL

JOURNAL OF CONTROL, AUTOMATION, AND SYSTEMS, PP.788-798.

[24] PEDRO F. FELZENSZWALB AND DANIEL P. HUTTENLOCHE, (2004) ” EFFICIENT GRAPH-BASED

IMAGE SEGMENTATION”, INTERNATIONAL JOURNAL OF COMPUTER VISION.

[25] YOUNG GEUN KIM AND HAKIL KIM, (2004) ”LAYERED GROUND FLOOR DETECTION FOR VISION

BASED MOBILE ROBOT NAVIGATION.”;IN IEEE ROBOTICS AND AUTOMATION (ICRA), VOLUME 1, PP 13 – 18.

[26] Y. J. JUNG, A. BAIK, J. KIM, AND D. PARK, (2009) ”A NOVEL 2D-TO-3D CONVERSION TECHNIQUE

BASED ON RELATIVE HEIGHT DEPTH CUE”; IN PROC. STEREOSCOPIC DISPLAYS AND APPLICATIONS XX, VOL.

7237.

[27] C.-C. CHENG, C.-T. LI, AND L.-G. CHEN,(2010) ” A NOVEL 2D-TO-3D CONVERSION SYSTEM USING

EDGE INFORMATION”; IEEE TRANS. CONSUMER ELECTRONICS, VOL. 56, NO. 3, PP. 1739–1745.

[28] R. C. GONZALEZ AND R. E. WOODS, (2010), DIGITAL IMAGE PROCESSING, THIRD EDITION. UPPER

SADDLE RIVER, NJ: PEARSON EDUCATION INC.

[29] W.-N. LIE, C.-Y. CHEN, AND W.-C. CHEN.(2011) “2D TO 3D VIDEO CONVERSION WITH KEY-FRAME

DEPTH PROPAGATION AND TRILATERAL FILTERING”; ELECTRON. LETT., VOL. 47, NO. 5, PP. 319–321.

[30] A. CRIMINISI, I. REID, AND A. ZISSERMAN,(2000) ”SINGLE VIEW METROLOGY. INTERNATIONAL

JOURNAL OF COMPUTER VISION”; 40:123148.

[31] FENG HAN, SONG-CHUN ZHU, (2004) ”AUTOMATIC SINGLE VIEW BUILDING RECONSTRUCTION BY

INTEGRATING SEGMENTATION”; CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION

WORKSHOP.


48

Figure 7. (a)First Frame (b) Intermediate Frame (c) Final Frame