Automatic Semantic Video Object Segmentation

Content-based video processing is one of the methods considered to meet the demands ofrnnewly emerging multimedia applications. For content-based processing, video has to bernsegmented into meaningful objects – semantic video objects. Many applications requirernautomatic segmentation (AS) of semantic video objects. However, AS is very challengingrntask.rnIn this thesis, an effective AS system is proposed by examining, selecting and combiningrnefficient and simplified techniques that are justified with theoretical analysis. The proposedrnsystem is developed and tested with the following three cases depending on whether therernexists camera motion or not in video sequences, and whether there is an initial backgroundrnreference frame. Case-1 is for video sequences with no camera motion and with initialrnbackground reference frame, Case-2 is for video sequences where there is no camera motionrnand no initial background reference frame, and Case-3 is for video sequences with camerarnmotion.rnChange detection is used as the main step in each of the cases to detect semantic objects andrnto produce object mask. Different problems in change detection like “uncoveredrnbackground”, “global motion of background (GMOB)” and “camera noise” are identified andrnsolved. Two change detection results are combined to remove the uncovered backgroundrnproblem. More emphasis is made for the problem of GMOB. A 3-level block-basedrnhierarchical motion estimation and affine parameter model for frame warping is used to solvernthis problem. Camera noise is removed by using model-based change detection.rnFor post-processing to improve the resulting change detection masks, a new filling-inrntechnique is proposed. This technique is used to fill open areas inside object regions withrnuniform intensity. To improve the boundary of segmentation masks, morphological openrnclose operations are used. The final semantic video objects are obtained by superimposing thernresulting mask over the original frame.rnxrnTest results show that the system effectively identified and segmented the semantic objects.rnSubjective evaluation of results for the three cases showed that among the window sizes usedrnin change detection, 5x5 and 7x7 produced better and comparable results in terms of visualrnquality and boundary smoothness. These results are obtained after applying one pair of openrnclose operation in Case-1 and two pairs in Case-2 and Case-3.rnBased on subjective comparison of results with other systems, 80% of observations for thernresults of Case-1 and 100% for Case-2 reported more pleasing results with smooth boundary.rnKeywords: content-based video processing, automatic segmentation, semantic video objects,rnchange detection, motion estimation, post-processing.

Get Full Work

Report copyright infringement or plagiarism