Simulation of video visual region of interest extraction based on eye tracking
Regions with significant features in videos are the region of interest.Because there are many single tar-get features in video images,the contour edge has a strong correlation,interfering with the extraction of the visual re-gion.Therefore,a method of extracting visual regions of interest in video was proposed.Firstly,we converted the color space of the original video in advance,thus reducing the influence of color.Secondly,we detected the edge of the orig-inal video,and used morphological close operation to fill the break of the video edge and thus remove noise.Thirdly,we used the rotation correction to get a simple and complete video image.Meanwhile,we used eye tracking technology to extract texture features and brightness saliency maps.Finally,we extracted the visual regions of interest.Experi-mental results show that the proposed method can accurately extract the visual region of interest,and the peak signal-to-noise ratio can reach up to 1.5dB,which verifies that the application accuracy and image quality gain of the meth-od are high.
Eye trackingEdge detectionRegion of interestMorphological close operationFeature extraction
NETL
NSTL
万方数据
