Research On ROI Segmentation In Stereo Video And Its Coding Technology

Stereo systems are designed to emulate human stereo perception,which have many applications,it is very popular and having wider application in this times that science and technology is rapid developing.Stereo vision infers three-dimensional(3-D) scene by generated recording two images with slightly different view points of the same scene. Compare with the normal vision,stereo vision has more quantities of data,in order to save the storage and bandwidth,efficient coding techniques should be employed to reduce the data rate.Block-based and Object-based methods are often used to coding stereoscopic image sequences.Block-based approach is simple and easy to achieve and allowing more straightforward hardware implementations,but the subjective quality of reconstructed images may be bad at low bit-rates.Object-based schemes alleviate the problem of annoying coding errors,providing a more natural representation of the scene,excellent performance and by producing fewer annoying effects such as blocking artifacts and mosquito effects. Therefore,it has been an active area of research as the developing trend of stereo video compression scheme.In ROI(Region of interest)-based coding,images should be segmented to several areas,the importance of which are different.Different compression strategies should be took according to the importance of each area.This method could not only maintaining the value of application of images,but also obtaining a higher Compression ratio.The storage and bandwidth could be saved as far as possible.ROI segmentation is a crucial step in object-based stereo video coding,accuracy of ROI segmentation takes a very important part of stereo video coding.Object segmentation algorithms are derived from image segmentation and one-channel video segmentation algorithms.Some researchers segment objects in one channel using traditional image segmentation algorithms,then get the objects of the other channel based on stereo matching. Some other researchers do the segmentation procedure on the depth map.This method tends to be more accurate,for the depth information is quite close to the true object boundary.This paper mainly works on the ROI segmentation and stereo video coding problem in stereo videos.Algorithms of disparity estimation are being studied and methods of image and video segmentation are also being realized.Background statistical technology and algorithm of background difference are being studied.The basic principles of H.264 and its process of encoding and decoding are also being understood.For ROI segmentation,an algorithm based on disparity estimation and background statistical technology is proposed. H.264 is employed to do stereo video coding.Steps of the algorithms above are described as follows:(1) The first segmentation procedure is based on the disparity map,for disparity map contains the depth information associated with the 3-D scene.One can get disparity map after stereo matching.For the parallel camera configuration,the epipolar lines are parallel to horizontal scan lines,so one can constrain the search within the horizontal scan lines.In this paper,we do histograms matching two times in order to reduce the influence of intensity differences between left and right image.First we match right histogram as left one is a standard,second match left one as right is a standard.Then we sub-sampling the images to decrease quantities of computing,match the individual pixels in one partition of the horizontal scan line in one image with the pixels in the corresponding partition of the scan line in the other image.We use a one-dimensional window to do the matching procedure. After computing a disparity map,information between scan lines is used to refine the disparity map.According to the coherence principle,image areas with smooth disparity variation belong to the same physical object.Then Canny edge detection is be employed to smooth the masks.(2) According to the coherence principle,image areas with smooth disparity variation belong to the same physical object.The regions which are discontinuity can let us get the initial object mask.Then over-sampling the mask to original size.The regions which are outside the object mask form the background.The whole background can be constructed from these regions of accumulated frames,compare the background with current frame,then we can segment the accurate target object mask,process the mask image using morphological methods,obtain the accurate object.Experimental results have shown the performance of the proposed scheme to be better on condition that images have stationary and complex background even if have more than one object in it.But the constructed backgrounds have a great influence on performance.(3) In this paper the international standard H.264 is employed to do the stereo video coding.Since the human eyes have higher visual resolution on the ROI region,therefore the bit-rate should be different between ROI region and background region,higher bit-rate is distributed to ROI regions and lower bit-rate is distributed to background regions in the main channel.For the vice channel,it does not need to be encoded,but do H.264 coding to disparity map between the left and right sequences.At last,vice channel sequence is computed by inverse operation of disparity estimation when decoding.For the main channel encoding bit-rate control method,first use a smaller quantify step size to do the ROI regions coding,then use coarse quantify step size to do coding of other regions.These two parameters of quantify could be gained by plus or minus a certain value to Q_p.In the coding process,If the encoding bit-rate is lower than the target bit-rate,Q_p,should be lowered to generate more bits,contrarily Q_p should be increase.ROI-based bit-rate control method can guarantee the quality of visual images,at the same time the coding bit-rate is similar to the target bit-rate.In comparison with objective parameters of the other methods,the algorithm which presented in this paper has a higher compression ratio in condition of get the nearly same PSNR.If the same compression ratio want to be get by other algorithms,only the coding bit-rate must be lower that it can be achieved,whereas that would result in the sharp decline in PSNR.The results show that algorithms proposed in this paper perform well in object segmentation in stereo videos.Meanwhile the algorithms are fast,and easy to implement. The coding algorithm proposed in this paper perform well in subjective effects,has higher PSNR and compression ratio.But the constructed backgrounds have a great influence on performance and our work is based on gray information of images and videos.In future work, we anticipate that by improving the disparity estimation algorithm,it should be possible to get more accurate disparity map.By adding motion information in the segmentation procedure,it will be promising to get more satisfying results.