The Research About Region Of Interest Extracyion Method Based On Saliency Map Of Video Image

The traditional visual attention models can extract the region of interest(ROI)better from the still image but when we apply these models on video, we can’t get themotion information. We process these images independently and ignore the linkagesbetween the images. Based on this situation, we add motion feature to improve thetraditional visual attention models and modeling novel visual attention model which iseasy to get saliency map from the video.The ROI extracting method based on video both ca lculate the saliency map fromdifferent space, so we have made full use of the motion information and refuse thelosing of motion information. We also improve the algorithm to calculate the saliencymap which is help to get the better ROI. We use spectral residual algorithm and PCTalgorithm to get the saliency map in this method, and we compare the advantages anddisadvantages of the two algorithms from the experiments. Finally, the better algorithmis finded.The model based on video saliency map is used to extract ROI. We use spectralresidual algorithm to get the saliency map, and we also add the motion information toget the motion saliency map. Then we get the improved saliency map, and we canextract ROI exactly. Original spectral residual algorithm only consider the intensityfeature, we add the color feature, direction feature and motion feature based on theoriginal algorithm. We use the frame difference method to acquire the motion feature.Combined with multi-scale analysis, the model is finished. Finally, some experimentresults is given to detect the performance of the model.