Video Object Segmentation Algorithm

The segmentation of moving objects in video sequences is very important for second generation coding and is a basic prerequisite for content-based video applications, which contain video searches, compression and object-oriented edition, aptitude human-machine exchange, etc. The results of objects segmentation will affect subsequent applications directly.At the present time, there is no current method, which can segment object models from the background efficiently, though a great deal of research work has been done for video coding.Based on research on methods for the segmentation of moving objects in video sequences, a new flexible method is proposed to achieve automatic segmentation from video sequences.First, moving regions are achieved through temporal segmentation, getting rid of making frame difference or twice frame difference image as models. The basic idea of the method is: frame mask image can be achieved with a threshold from moving information got by intersections of twice frame difference images. In order to make a complete result, moving regions with part of inexact edges can be reached by twice scan and morphological operations. The threshold required by the threshold procedure can be got by LHS (least half samples) method. With LHS method, we get a threshold in time. And this is a key part for automatic segmentation.Then, a morphological gradient operator is used for spatial segmentation to find edges of images. After being processed, intensities of pixels near objects edges are upgraded, while intensities of other pixels are suppressed. So, the edges are extracted completely.At last, combining temporal and spatial segmentation can remove the redundant part of moving regions. Then mask of moving objects, got by twice scan, is filled to be a segmented moving object.The emphasis of this method is on temporal segmentation, which avoids use ofcommon methods with a high computational load. It is superior in reducing complexity and runtime, increasing currency and flexibility. In general, it has characters as follows:Firstly, flexibility is an outstanding character. According to the varying complexity of the background in video sequences, we can choose part or all of this method to realize segmentation.For sequences with a simple background, it is feasible to use only temporal segmentation. Even so, moving regions can be got flexibly by "AND" of n twice frame difference images. The value of n is flexible (?=l,2,3*'O- According to the results of simulations, it is noticed that, the method is optimal for results and time superiority when n equals 3.As for sequences with a complex background, the method can guarantee the results with good quality.Secondly, this method has simple logic and short runtime. Temporal segmentation is achieved by the idea of change-detection. Moving regions are extracted exactly by "AND" of twice frame difference images. At the same time spatial segmentation reduces computation complexity and runtime, edges from spatial segmentation can ensure the quality of segmented objects.Thirdly, better currency for head-shoulder sequences. Twice scan ensures integrity of mask image. This article contains many head-shoulder sequences to examine the results, which are satisfactory.Plenty of simulations have been carried out by this segmentation method. And results show that this method is simple and feasible. It has greater segmentation speed, better results and stronger robustness to noise.