Research On Saliency Detection Technology For Integrated Imaging Light Field View

Compared with other stereoscopic display technologies,integrated imaging technology has many advantages,which is an important trend of stereoscopic display in the future.At present,the main problems faced by the integrated imaging technology are the low resolution of the 3D reconstructed image,the serious graininess felt when viewing,and the stereo effect is not very satisfactory.This paper mainly researches on the saliency detection technology for the integrated imaging light field view,using the saliency detection technology to extract the saliency objects in the light field view captured by the integrated imaging camera array,and provide a good source of saliency targets for the subsequent 3D model reconstruction step of the stereo content generation part.However,the existing saliency detection methods do not perform well in detail when dealing with certain complex real scenes,such as scenes containing multiple salient targets or low contrast between some regions of the salient targets and the background region,and there will be some problems such as mistaking the background region as the saliency region or missing the saliency region.Therefore,this paper proposes a saliency detection algorithm combined with optical field refocusing.Firstly,a specified depth layer in the center view image can be refocused through multiple view images of the scene,and the objects on the depth layer can be clearly displayed,while the objects on other depth layers are blurred to varying degrees.Then,taking the refocused image as the input of saliency detection,the saliency detection based on graph is carried out by using the depth and focus information of the scene.On the basis of the graph based manifold ranking saliency detection method,the global and local smoothness constraints are proposed to prevent the propagation of error labels,and further improve the fineness of detection results.The experimental results on 4D light field image data set show that the saliency detection results of the proposed algorithm are closer to the artificially labeled ground truth,and the mean absolute error between saliency map and ground truth is lower than that of the existing methods.In addition,this algorithm is used to detect the saliency of multi view image array captured from real scene,and the detection results contain abundant information of saliency target.The main research work of this paper is as follows:1.Research on the principle of integrated imaging technology.The realization process of integrated imaging technology for true 3D stereoscopic display is studied from two parts of light field capture process and optical reconstruction process.In the acquisition part,the sparse camera array is used to capture the real scene.In the stereo content generation part,the 3D model is reconstructed and the virtual viewpoint image is generated by the computer,and then the stereo element image array for display on the stereo display system is synthesized.The related technologies involved in the technical route are studied,which provides a clear understanding of the research direction and goal of saliency detection technology for integrated imaging light field view.2.Research on image refocusing algorithm based on light field.The light field capture system of camera array is built,and the principle of computational refocusing based on multi view image array is studied.Aiming at the problem of low image quality after refocusing caused by multiple interpolation,a selective optical field view refocusing algorithm based on super-resolution cooperative reconstruction is proposed.The algorithm integrates the bokeh rendering and super-resolution reconstruction into the same scheme to generate the refocused image.The experimental results show that for the input multiple view images,by selecting a depth layer of the scene for refocusing,the objects on the depth layer can be clearly displayed,and the objects on other depth layers can be blurred in varying degrees.3.Research on saliency detection algorithm based on graph.When the existing saliency detection methods deal with the images captured by the camera array from the complex real scene,the saliency map is not fine enough,and even some saliency regions will be lost.In this paper,we propose a saliency detection method based on selective light field refocusing of multi view image array.Through multiple view images of the same scene,we use the depth,focus and other information of the scene to combine with the graph based saliency detection method.At the same time,we propose to combine the global and local smoothness constraints to solve the above problems.The experimental results show that the proposed algorithm has good effect,compared with other methods,the saliency map is the most precise,and the mean absolute error between saliency map and truth map is reduced.