Efficient Moving Object Segmentation Algorithm
Using Background Registration Technique
Shao-Yi Chien, Shyh-Yih Ma, and Liang-Gee Chen, Fellow, IEEE
Hsin-Hua Lee
IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY, JULY 2002
Outline
•
•
•
•
•
Introduction
Segmentation Algorithm
Shadow Effects
Experimental Results
Conclusion
Introduction
• Video segmentation is a key operation for contentbased video coding.
• For example, MPEG-4 enables the content-based
functionalities by using VOP (video object plane) as
the basic coding element.
• The authors propose an efficient algorithm suitable
for real-time content-based multimedia
communication system.
Introduction (Cont.)
• Conventional video segmentation algorithm can be
roughly classified into two categories by their primary
segmentation criteria.
• spatial homogeneity
• Track the object boundary more precisely than other
methods, but computation complexity is very high.
• change detection
Change detection
•
Conventional main steps
1.
2.
•
Position and shape of the moving object is detect from the
frame difference of two consecutive frames.
Boundary fine-tuning process based on spatial and
temporal information.
It’s thought that these approach is more efficient than the
previous category because it is the motion that
distinguishes a moving object from the background.
Segmentation Algorithm
• The basic idea of the proposed segmentation
algorithm is change detection.
• The authors construct and maintain up-to-date
background information from the video sequence and
compare each with the background.
• Any pixel that is significant different from the
background is assumed to be in the object region.
Segmentation Algorithm (Cont.)
Frame Difference
• Stationary background
the characteristics is well known and more reliable.
• Long-term behavior
the object motion accumulated from several frames instead
of relying on frame difference of two consecutive frames
only.
Frame Difference
Frame difference mask.
(a)(c) The original image. (b)(d) Frame difference mask
Background Registration
• Construct a reliable background information
• Maintain Stationary Map
If the value in the stationary map exceeds a
predefined value, then the pixel value in the current
frame is copied to the corresponding pixel in the
background buffer.
• The value in the background registration mask
indicates that whether the background information of
the corresponding pixel exists or not.
Background Registration (Cont.)
Construction and updating of the background buffer.
(a)(c) Original frame (b)(d) Constructed background
Background Difference
• Generates a background difference mask by
thresholding the difference between the current frame
and the background information stored in the
background buffer.
Object Detection
Post Processing
• Remove noise regions and to smooth the object
boundary.
• Small region filtering
• Close-open operation
Post Processing
Effect of noise region elimination.
(a) Mask after small-region filtering.
(b) Final object mask after close-open operation.
Post Processing (Cont.)
(c) Initial object mask.
(d) Final object mask after the noise region elimination step.
Post Processing (Cont.)
(e) Original image. (f) Segmented object.
Shadow Effects
• In situations where object shadows appear in the
background region, a pre-processing gradient filter is
applied on the input image to reduce the shadow
effect.
Shadow Effects (Cont.)
Shadow Effects (Cont.)
Effect of gradient filter. (a) Original image. (b) Segmentation result of the
original image. (c) Gradient image after applying the morphological
gradient operation. (d) Segmentation result of the gradient image.
Experimental Results
(Objective Evaluation )
Error rate in each frame of the Weather sequence (CIF).
Experimental Results
(Subjective Evaluation )
Experimental Results
(Subjective Evaluation ) (Cont.)
Conclusion
• In this paper, the author proposed an efficient moving
segmentation algorithm by avoiding the use of
computation intensive operations.
• The experimental results demonstrate that good
segmentation quality can be obtained efficiently;
therefore, this algorithm is very suitable for the realtime VOP generation in MPEG-4 multimedia
communication systems.