VIDEO SEGMENTATION FOR MOVING OBJECT DETECTION USING LOCAL CHANGE & ENTROPY BASED ADAPTIVE WINDOW THRESHOLDING

Natarajan Meghanathan et al. (Eds) : ITCSE, ICDIP, ICAIT - 2013
pp. 155–166, 2013. © CS & IT-CSCP 2013 DOI : 10.5121/csit.2013.3916
VIDEO SEGMENTATION FOR MOVING
OBJECT DETECTION USING LOCAL
CHANGE & ENTROPY BASED ADAPTIVE
WINDOW THRESHOLDING
Anuradha.S.G 1
, Dr.K.Karibasappa 2
, Dr.B.Eswar Reddy 3
1
Department of Computer Science & Engineering, RYMEC, Bellary, Karnataka,
India.
anuradha_gagadin@rediffmail.com
2
Department of Information Science & Engineering, DSCE, Bangalore, India.
K_karibasappa@hotmail.com
3
Department of Computer Science & Engineering,
JNTUA, Anantapur, Andhra Pradesh, India
eswarcsejntu@gmail.com
ABSTRACT
Motion detection and object segmentation are an important research area of image-video
processing and computer vision. The technique and mathematical modeling used to detect and
segment region of interest (ROI) objects comprise the algorithmic modules of various high-level
techniques in video analysis, object extraction, classification, and recognition. The detection of
moving object is significant in many tasks, such as video surveillance & moving object tracking.
The design of a video surveillance system is directed on involuntary identification of events of
interest, especially on tracking and on classification of moving objects. An entropy based real-
time adaptive non-parametric window thresholding algorithm for change detection is
anticipated in this research. Based on the approximation of the value of scatter of sections of
change in a difference image, a threshold of every image block is calculated discriminatively
using entropy structure, and then the global threshold is attained by averaging all thresholds for
image blocks of the frame. The block threshold is calculated contrarily for regions of change
and background. Investigational results show the proposed thresholding algorithm
accomplishes well for change detection with high efficiency.
KEYWORDS
Video Segmentation, Object Detection, Object Tracking, Video Surveillance, motion detection.
1. INTRODUCTION
The development of video databases has impelled research for structuring multimedia content.
Traditionally, low-level descriptions are provided by image and video segmentation techniques.

156 Computer Science & Information Technology (CS & IT)
The best segmentation is achieved by the human eye, performing simultaneously segmentation
and recognition of the object thanks to a strong prior knowledge about the objects’ structures. To
generate similar high-level descriptions, a knowledge representation should be used in computer-
based systems. One of the challenges is to map efficiently the low-level descriptions with the
knowledge representation to improve both segmentation and interpretation of the scene.
There are three key steps in video analysis: detection of interesting moving objects, tracking of
such objects from frame to frame, and analysis of object tracks to recognize their behaviour.
In its simplest form, Segmentation of moving objects in image sequences is one of the key issues
in computer vision, since it lies at the base of virtually any scene analysis problem. In particular,
segmentation of moving objects is a crucial factor in content- based applications such as
interactive TV, content-based scalability for video coding, content-based indexing and retrieval,
etc. Obviously, such applications require an accurate and stable partition of an image sequence to
semantically meaningful objects.
2. PROBLEM & RELATED WORK
Commonly, two approaches exist for detecting moving objects in a sequence of images. The first
approach relies on change detection as the source of temporal information. These approaches are
motivated by the assumption that moving objects usually entail intensity changes between
successive frames. Hence, moving objects can be identified by applying a decision rule on the
intensity differences between successive frames in a sequence. Various decision rules have been
suggested to identify moving objects. Yet, these approaches suffer from two major drawbacks.
First, unless moving objects are sufficiently textured, only occlusion areas (covered background)
will be marked as changed, while the interior of objects will remain unchanged.
Second, uncovered background will be marked as changed in the process as well, thus the
boundaries of the moving objects are likely to be inaccurate. To overcome this drawback, a post-
processing step that distinguishes between moving objects and uncovered background has to be
applied.
The aim of an object tracker is to generate the trajectory of an object over time by locating its
position in every frame of the video. Object tracker may also provide the complete region in the
image that is occupied by the object at every time instant. The tasks of detecting the object and
establishing correspondence between the object instances across frames can either be performed
separately or jointly.
Depending on the tracking domain, a tracker can also provide object-centric information, such as
orientation, area, or shape of an object. One can simplify tracking by imposing constraints on the
motion and/or appearance of objects.
The basic framework of moving object detection for video surveillance is shown in figure below.

Computer Science & Information Technology (CS & IT) 157
Figure1. Framework for Basic Video Object Detection System
In computer vision and video processing areas, moving object detection is a very important
research topic. The process of moving object detection in video consists of two steps -
background extraction and moving object detection. The preliminary idea is to capture a series of
video pictures at regular intervals; the video is divided into n number of frames to describe the
vector information of the region.Rajaguru et al. in [40], develops a work in Moving object
detection using draw pad, experiments show that detected objects in real streaming video is
converted into drawing shapes or figures in a specified area. A paper [41] begins with a
discussion of the deficiency of current algorithms which use the number of the changing points of
frame-difference as the threshold to determine a moving object.
A number of techniques and algorithms have been developed for the object segmentation task in
the past [1-5]. The technique presented in certain researches segments moving objects directly
relied on motion information. The affine parametric clustering approach provided a layered
description of video by its motion similarity depicted in [1]. The technique in [2] combined
motion estimation and segmentation within the Bayesian framework. The interdependence
between the optical flow estimation and the segmentation map has been reinforced iteratively. In
[3], an integrated Bayesian approach was proposed to extract motion layers. Yet, most physical
objects usually exhibit non-rigid motion and cannot be characterized by a parametric model, thus
motion-based moving object segmentation techniques result in a finer partition than actually
requirement.
In contrast with the motion-based approach, region-based object segmentation technique is
developed to accommodate the accurate segmentation of objects [4-5]. The spatial partition of
frame is firstly performed to obtain homogeneous regions. The location of object boundary is
guided by the initial spatial segmentation. Therefore, the extracted objects can succeed in object
boundaries that are not clearly distinct. In [4], the first frame of a video sequence is partitioned
into homogeneous regions based on intensity. A region merging process grouped the regions into
object based on motion similarity. In order to get meaningful spatial partition, human interaction
and the morphological segmentation tool was combined to obtain a complete and accurate
extraction of video object in the first frame [5].
Level set methods are also important in problems with current scenario, in a research [42], firstly,
an improved frame difference algorithm is used for dealing with the problem of initial moving
objective detection in video sequence, with adjacent video frame subtracting technique, moving

objective pixels in current image can be detected by adaptive threshold. Then expanding level set
C-V model, the contours of moving object should be extracted with the energy function evolving.
An efficient Modified Directional Lifting-based 9/7 Discrete Wavelet Transform (MDLDWT)
structure is proposed in [43] to further reduce the computational cost and preserve the fine shape
information in low resolution image.
The most challenging part of moving objects detection in moving background is that motion
induced by camera moving may dominate the observed motion, which makes existing methods
cannot detect the real moving objects from the moving background robustly and computationally
efficient. Xie, Songhua et al in [44], proposed a very fast method that can detect moving objects
from moving background accurately without prior knowledge of the camera motion
characteristics. Also, a Multi-layer Codebook-based Background Subtraction (MCBS) model is
proposed in [45] for video sequences to detect moving objects.
Turbulence mitigation refers to the stabilization of videos with non uniform deformations due to
the influence of optical turbulence. Typical approaches for turbulence mitigation follow
averaging or dewarping techniques. Although these methods can reduce the turbulence, they
distort the independently moving objects, which can often be of great interest, [46] address the
novel problem of simultaneous turbulence mitigation and moving object detection.
Moving object detection is an imperative constituent of a traffic surveillance structure. Usual
background subtraction methodologies often poorly achieve on a long out-of-doors traffic video
due to vehicles waiting at an intersection & gradual variations of illumination and background
shadow location. [47] present a fast and vigorous background subtraction algorithm established
on unified spatio-temporal background and foreground demonstration. The connection between
neighbouring pixels provides high levels of detection accurateness in the dynamic background
scene. Author’s method (Bayesian fusion), which establishes the traffic flow scene model,
associates both background & foreground models and deliberates prior probabilities to adapt
changes of background in every frame.
There has been a increasing research curiosity in video image segmentation over the past period
& towards this end, a wide variability of procedures have been established [18]-[21].
The video segmentation methodologies have extensively used stochastic image models,
particularly Markov Random Field (MRF) model, as the model for video sequences [22]-[24].
MRF model has proved to be an effective stochastic model for image segmentation [25]-[27]
because of its attribute to model context dependent entities such as image pixels and correlated
features. In Video segmentation, besides spatial modelling and constraints, temporal constraints
are also added to devise spatio-temporal image segmentation schemes.
An adaptive clustering algorithm has been reported [32] where temporal constraints and temporal
local density have been adopted for smooth transition of segmentation from frame to frame.
Spatio-temporal segmentation has also been applied to image sequences [28] with different
filtering techniques. Extraction of moving object and tracking of the same has been achieved in
spatio-temporal framework [29] with Genetic algorithm serving as the optimization tool for
image segmentation.
Recently, MRF model has been used to model spatial entities in each frame [29] and Distributed
Genetic algorithm (DGA) has been used to obtain segmentation. Modified version of DGA has

been proposed [23] to obtain segmentation of video sequences in spatio-temporal framework.
Besides, video segmentation and foreground subtraction has been achieved using the spatio-
temporal notion [30]-[31] where the spatial model is the Gibbs Markov Random Field and the
temporal changes are modelled by mixture of Gaussian distributions. Very recently, automatic
segmentation algorithm of foreground objects in video sequence segmentation has been proposed
[32].
3. THE PROPOSED ALGORITHM
The proposed algorithm computes a threshold of each block of an image adaptively based on the
scatter of regions of change (ROC) using the local change adaptive thresholding and averages all
thresholds for image blocks to obtain the global threshold based on entropy values. To enhance
the segmentation results, the results thus obtained from ROC thresholding section are verified and
compensated by considering the information of the pixels belonging to objects in the previous
frame. After deciding the window size entropy of the window is computed from the gray level
distribution of the window and background subtraction (non ROC region) is done after
calculation of entropy.
Entropy-based methods consequence in algorithms that use the entropy of the foreground and
background sections, the cross-entropy between the original and binarized image, etc. The
entropy of a system as defined by Shannon gives a measure of our ignorance about its actual
structure.
One of the primary disturbances sources are uneven lighting, which often exists in the capturing
of an image, especially during field operation. The main causes for uneven lighting are:
1. The light may not be always stable
2. The object is so large such that it creates an uneven distribution of the light, and
3. The background is unable to be optically isolated from shadows of other objects.
One possible solution to this problem is to partition the whole image into certain small windows,
and then use those existing methods to threshold each small window. This process is called
thresholding in partitioned windows. The smaller the window size is, the better the result will be.
However, when the window size becomes too small, it can produce the problem of homogeneous
windows, i,e., windows contain only background or object pixels. As a consequence, black areas
called ghost objects will occur after thresholding. Therefore, there is a need to develop a new
technique for automatically selecting window size in order to obtain optimal result i.e., adaptive
window selection. This technique is based on the pyramid data structure manipulation, and the
window size is adaptively selected according to Shannon’s entropy measure.
If a global thresholding algorithm is applied in a difference image, affected by noise,
illumination variation or shading:
i. Few pixels that actually correspond to the background in an image frame are identified as
changed pixels, whereas these are actually not.
ii. It also happens that a pixel in the difference image that correspond to actual change
region and lies in the lower range of the histogram may be identified as an unchanged
pixel.

An adaptive thresholding approach can be used to overcome these problems. However the choice
of window size is an issue.
In this regards we propose entropy based adaptive window selection scheme to determine the
block/ window size in background segmentation process. Here the threshold value for a particular
window is obtained by Otsu’s thresholding scheme. To enhance the segmentation results, the
results thus obtained from ROC segmentation are verified and compensated by considering the
information of the pixels belonging to objects in the previous frame. This is represented as
Where, R is a matrix having the same size of the frame, s is the element number in the matrix and
rs is the value of the VOP at location s. If a pixel is found to have rs = 1, it is a part of the moving
object of the previous frame; otherwise it belongs to the background of the previous frame.
Based on this information, Thresholding in Background (Non ROC) is modified as follows: if it
belongs to a moving object part in the previous frame and its label obtained by temporal
segmentation is the same as one of the corresponding pixels in the previous frame, the pixel is
marked as the foreground area in the current frame else as a background.
A. Thresholding In ROC
Since size and velocity of objects, noise, local changes in videos may affect the histogram of a
wk, the first moment of histogram is used to estimate the scatter of ROC making the estimation
(Eq.4 and 5) adaptive to these characteristics. We also account for local changes (Eq. 6).
The ROC in Dn are, in general, scattered over the K image blocks. Let be a pixel in Dn that
varies between 0 and 255. i is high in ROC which are caused by strong changes such as motion or
significant illumination changes and is low in non ROC which are caused by slight changes such
as noise or slight illumination changes. We use the first moment, mk, of the histogram of each
image block wk as a measure for determining if an image block contains ROC.

Where is the frequency of Gray-level i= {1, 2, •••, gmax} and gmax is the maximum Gray-level in
the image block. If of is greater than a threshold , the image block is regarded as a block
containing ROC, and marked as , otherwise, it is marked as , i.e.,
May vary greatly under different video conditions. We propose, therefore, to adaptively
determine as follows. Non-zero pixels in image blocks without ROC are in general caused by
noise, illumination changes, or background movement. Those factors usually affect an image
globally thus mk for different should be similar to each other. To find , we first compute the
mk of each block and then descending sort them k values. A straight line between the first bin and
the last filled bin is then drawn. is selected to maximize the perpendicular distance between the
line and the sorted first moment curve.
For each image block containing ROC, we propose the local-change adaptive thresholding.
First, the histogram of each is computed and divided into L equal partitions, and the most
frequent gray-level , l = {1, 2, ••• , L}, in each histogram partition is fixed. Then the average
Gray level of is computed and the threshold of is obtained by averaging the sum of
all and , i.e.,
B. Thresholding in background (Non ROC)
The basic notion of window growing is to fix the window size primarily focussing on the
information measure of the image at different scale. In other words, fixing the size of the window
not only depends on the entropy of the chosen window but also the feature entropy of the
window. The edges of the window are considered as the features and the feature entropy is
computed. Since, the edge map represents the image information at a different scale, the entropy
at this scale also plays a pivotal role for image segmentation. Thus, the basic notion is to capture
the information at a different scale. In this approach an arbitrarily small window (here the
window size w is chosen as 5 × 5) is considered initially and the entropy of the window is
computed from the gray level distribution of the window and is denoted by Hw
Where the probability distribution of the ith
Gray value is, denotes entropy of the window, G
denotes the total number of Gray values. Over a given window, the edge map is computed and the
entropy of the edge map is

Where denotes the entropy of the edge map of the window. The following are the two
window growing criterion.
If the entropy of the window is comparable to some fraction of the entropy of the whole image,
(represented as Th), that window is chosen for segmentation (by Otsu’s thresholding); otherwise
the window will be incremented by ∆w (here ∆w is considered 2) and the condition will be tested
again. The window will be fixed if the total image is exhausted. The final threshold image is
obtained by taking union of each considered threshold windows.
The window is fixed if the following is satisfied
Where Th is selected based on the entropy of the total image.
The following criterion is considered for window fixing after grow of the window.
Subject to the constraint,
4. RESULTS
We have modified the thresholding step of this algorithm since often leads to a significant level of
miss classifications.
Figure 2 Two regions (in bounding boxes) of an image

Figure 3 Initial Adaptive Thresholding on interconnected basis for the original input frame
Figure 4 Tm is selected to maximize the perpendicular distance between the line and the sorted first
moment curve
Figure 5. Input frame and segmented image by the user

(a) (b)
(c) (d)
Figure 6. (a) Shows the original frame content in a video stream, (b) and (c) shows foreground abstraction
through adaptive thresholding and window based thresholding (d) shows final segmentation through
entropy based threshold selection
Figure 7. Detection failure characteristics for different false alarm for input streams
5. CONCLUSIONS
Real-time entropy based adaptive window thresholding method for video object change detection
is proposed in this research. Local changes are taken into account. Based on the estimation of the
scatter of regions of change of a difference image, a threshold of each image block is computed
discriminatively with the help of entropy scheme. The universal threshold is the average of all
thresholds of image blocks. Both the visual assessment and the objective estimation show that the
proposed method clearly outperforms the other traditional methods and is steadier than the spatial
properties based methods. The efficiency of the proposed method is much sophisticated than the
spatial properties based methods.

REFERENCES
[1] M.M.Chane. kM.Tekalo. andM.I.Sezan. “Simultaneous motion estimation and segmentation,” IEEE
Transaction on image Processing, VoL6, No. 9, pp. 1326-1333, Sept. 1997.
[2] J.Wangand E.Adelson, ‘‘ Representing moving images with laver.” IEEE Transaction on Image
Processing. Vo1.3. NoS.DD. 6i5-638, Sept. 1994.
[3] P.H.S.Torr, RSzeliski, and P.Anandan,”An integrated Bayesian approach to layer extraction from
image sequence,” IEEE Transaction on Pattern Analysis and Machine Intelligence, Vol. 23, No.3, pp.
297-303, Mar. 2001.
[4] D. Wang, ”Unsupervised video segmentation based on watersheds and temporal tracking,” IEEE
Transaction on Circuits and Systems for Video Technology, Vol. 8, No.5, Sept. 1998
[5] C.Gu and M.C.Lee, ”Semiautomatic segmentation and tracking of semantic video objects,”IEEE
Transaction on Circuits and Systems for Video Technology, Vo1.8, No.5, Sept 1998.
[6] Hiraiwa, A., Accurate estimation of optical flow for fully automated tracking of moving-objects
within video streams, Circuits and Systems, 1999. ISCAS '99. Proceedings of the 1999 IEEE
International Symposium
[7] Takaya, K., Tracking a video object with the active contour (snake) predicted by the optical flow,
Electrical and Computer Engineering, 2008. CCECE 2008. Canadian Conference
[8] Jayabalan, E., Non Rigid Object Tracking in Aerial Videos by Combined Snake and Optical flow
Technique, Computer Graphics, Imaging and Visualisation, 2007. CGIV '07
[9] Bauer, N.J., Object focused simultaneous estimation of optical flow and state dynamics, Intelligent
Sensors, Sensor Networks and Information Processing, 2008. ISSNIP 2008. International Conference
[10] Ping Gao, Moving object detection based on kirsch operator combined with Optical Flow, Image
Analysis and Signal Processing (IASP), 2010 International Conference
[11] Pathirana, P.N., Simultaneous estimation of optical flow and object state: A modified approach to
optical flow calculation, Networking, Sensing and Control, 2007 IEEE International Conference
[12] Tiesheng Wang, Moving Object Tracking from Videos Based on Enhanced Space-Time-Range Mean
Shift and Motion Consistency, Multimedia and Expo, 2007 IEEE International Conference
[13] Khan, Z.H., Joint particle filters and multi-mode anisotropic mean shift for robust tracking of video
objects with partitioned areas, Image Processing (ICIP), 2009 16th IEEE International Conference
[14] YuanyuanJia, Real-time integrated multi-object detection and tracking in video sequences using
detection and mean shift based particle filters, Web Society (SWS), 2010 IEEE 2nd Symposium
[15] Sung-Mo Park, Object tracking in MPEG compressed video using mean-shift algorithm, Information,
Communications and Signal Processing, 2003 and Fourth Pacific Rim Conference on Multimedia.
Proceedings of the 2003 Joint Conference of the Fourth International Conference
[16] Khan, Z.H., Joint anisotropic mean shift and consensus point feature correspondences for object
tracking in video, Multimedia and Expo, 2009. ICME 2009. IEEE International Conference
[17] Yilmaz, A., Object Tracking by Asymmetric Kernel Mean Shift with Automatic Scale and
Orientation Selection, Computer Vision and Pattern Recognition, 2007. CVPR '07. IEEE Conference
[18] A. M. Teklap, Digital Video Processing. Prentice Hall, NJ, 1995.
[19] P. Salember and F. Marques, “Region based representation of image and video segmentation tools for
multimedia services,”IEEE Trans. Circuit systems and video Technology, vol. 9, No. 8, pp. 1147-
1169, Dec.1999.
[20] E. Y. Kim, S. H. Park and H. J. Kim, “A Genetic Algorithm-based segmentation of Random Field
Modeled images,” IEEE Signal processing letters, vol. 7, No. 11, pp. 301-303, Nov. 2000.
[21] S. Geman and D. Geman, “Stochastic relaxation, Gibbs distributions and the Bayesian restoration of
images,” IEEE Trans. on Pattern Analysis and Machine Intelligence, vol. 6, No. 6, pp. 721-741, Nov.
1984.
[22] R. O. Hinds and T. N. Pappas, “An Adaptive Clustering algorithm for Segmentation of Video
Sequences,” Proc. of International Conference on Acoustics, speech and signal Processing, ICASSP,
vol. 4, pp. 2427-2430, May. 1995.
[23] E. Y. Kim and K. Jung, “Genetic Algorithms for video Segmentation,” Pattern Recognition, vol. 38,
No. 1, pp.59-73, 2005.

[24] E. Y. Kim and S. H. Park, “Automatic video Segmentation using genetic algorithms,” Pattern
Recognition Letters, vol. 27, No. 11, pp. 1252-1265, Aug. 2006.
[25] Stan Z. Li, Markov field modeling in image analysis. Springer: Japan, 2001.
[26] J. Besag, “on the statistical analysis of dirty pictures,” Journal of Royal Statistical Society Series B
(Methodological), vol. 48, No. 3, pp.259-302, 1986.
[27] A. L. Bovic, Image and Video Processing. Academic Press, New York, 2000.
[28] G. K. Wu and T. R. Reed, “Image sequence processing using spatiotemporal segmentation,” IEEE
Trans. on circuits and systems for video Technology, vol. 9, No. 5, pp. 798-807, Aug. 1999.
[29] S. W. Hwang, E. Y. Kim, S. H. Park and H. J. Kim, “Object Extraction and Tracking using Genetic
Algorithms,” Proc. of International Conference on Image Processing, Thessaloniki, Greece, vol.2, pp.
383-386, Oct. 2001.
[30] S. D. Babacan and T. N. Pappas, “Spatiotemporal algorithm for joint video segmentation and
foreground detection,” Proc. EUSIPCO, Florence, Italy, Sep. 2006.
[31] S. D. Babacan and T. N. Pappas, “Spatiotemporal Algorithm for Background Subtraction,” Proc. of
IEEE International Conf. on Acoustics, Speech, and Signal Processing, ICASSP 07, Hawaii, USA,
pp. 1065-1068, April 2007.
[32] S. S. Huang and L. Fu,” Region-level motion-based background modeling and subtraction using
MRFs,” IEEE Transactions on image Pocessing, vol.16, No. 5, pp.1446-1456, May. 2007.
[33] Comaniciu, D., Meer, P., “Mean shift: A robust approach toward feature space analysis”, IEEE
Transactions on Pattern Analysis and Machine Intelligence, vol. 24, no. 5, pp. 603–619, 2002.
[34] Shan, C., Tan, T., Wei, Y., "Real-time hand tracking using a mean shift embedded particle filter",
Pattern Recognition, Vol. 40, No. 7, pp. 1958-1970, 2007.
[35] Yilmaz, A., Javed, O., Shah, M., “Object tracking: A survey”, ACM Computing Surveys, Vol. 38,
No. 4, Article No. 13 (Dec. 2006), 45 pages, doi: 10.1145/1177352.1177355, 2006.
[36] Comaniciu, D., Meer, P., "Mean Shift Analysis and Applications", IEEE International Conference
Computer Vision (ICCV'99), Kerkyra, Greece, pp. 1197-1203, 1999.
[37] Kass, M., Witkin, A., and Terzopoulos, D., “Snakes: active contour models”, International Journal of
Computer Vision, Vol. 1, No. 4, pp. 321–331, 1988.
[38] T. Bernier, J.-A. Landry, "A New Method for Representing and Matching Shapes of Natural Objects",
Pattern Recognition 36, 2003, pp. 1711-1723.
[39] H. Blum, “A Transformation for extracting new descriptors of Shape”, in: W. Whaten-Dunn(Ed.),
MIT Press, Cambridge, 1967. pp. 362-380.
[40] Rajaguru, D., Moving object detection using draw pad, Advances in Engineering, Science and
Management (ICAESM), 2012 International Conference
[41] HaihuiGao, A New Detection Algorithm of Moving Objects Based on Human Morphology,
Intelligent Information Hiding and Multimedia Signal Processing (IIH-MSP), 2012 Eighth
International Conference
[42] Rongguo Zhang, Moving Objective Detection and Its Contours Extraction Using Level Set Method,
Control Engineering and Communication Technology (ICCECT), 2012 International Conference
[43] Chih-Hsien Hsia, Improved directional lifting-based discrete wavelet transform for low resolution
moving object detection, Image Processing (ICIP), 2012 19th IEEE International Conference
[44] Xie, Songyun, Fast detecting moving objects in moving background using ORB feature matching,
Intelligent Control and Information Processing (ICICIP), 2013 Fourth International Conference
[45] Juo, J., Fast Background Subtraction Based on a Multi-Layer CodeBook Model for Moving Object
Detection, Circuits and Systems for Video Technology, IEEE Transactions

VIDEO SEGMENTATION FOR MOVING OBJECT DETECTION USING LOCAL CHANGE & ENTROPY BASED ADAPTIVE WINDOW THRESHOLDING

More Related Content

What's hot (18)

Similar to VIDEO SEGMENTATION FOR MOVING OBJECT DETECTION USING LOCAL CHANGE & ENTROPY BASED ADAPTIVE WINDOW THRESHOLDING (20)

More from cscpconf (20)

Recently uploaded (20)

VIDEO SEGMENTATION FOR MOVING OBJECT DETECTION USING LOCAL CHANGE & ENTROPY BASED ADAPTIVE WINDOW THRESHOLDING