Colorimetric Compensation Technique For Image Acquisition By Virtual Micro-lenses

In recent years,integrated imaging technology has attracted more and more attention from the industry as a new type of three-dimensional imaging and display technology with continuous viewing angle,full parallax,and no need for auxiliary viewing equipment.In order to overcome the disadvantages of traditionally collecting three-dimensional scene information through a lens array,as well as providing a variety of real large scenes and high resolution image sources for integrated imaging research,method for constructing virtual micro-lens array image by using multi-viewpoint image acquisition means and virtual viewpoint rendering algorithm has important practical significance and academic value.Virtual micro-lens capture images' color compensation can be achieved by correcting the color of the synthesized virtual viewpoint.Referring to the rendering of virtual viewpoints,this paper proposes a virtual viewpoint rendering algorithm based on image layering.First,perform a local optimal stereo matching on the reference viewpoint image and roughly extract the depth information,layered the reference images and calculated the parallax based on depth information.Then,the edges of the layered image are extended to reduce the gaps generated by the overlay,weighted translation of layered images according to the relative positions of virtual and reference viewpoints.Finally,according to the order of the foreground after the background,the virtual viewpoint is rendered layer by layer,and the virtual viewpoint image at any point within the 2(H)×2(V)viewpoint can be obtained.Comparing with the main algorithm for virtual point of view rendering DIBR,this method does not need to provide depth information in advance,nor does it have a high degree of dependence on the accuracy of the depth map.The mapping operation is relatively simple and has high imaging quality.Referring to color compensation,in order to make the rendering effect and color of the virtual viewpoint image closer to the real viewpoint image,this paper builds a constraint factor based on Euclidean distance for layered images,making the virtual viewpoint color not a coverage of a certain viewpoint,but a weighted sum of the color information of the four reference viewpoints.In addition,in order to obtain a color effect that is more in line with human eyes when stereoscopically displayed,perform color enhancement on the extended dense viewpoint image with brightness-enhanced world grayscale method based on image entropy.Under the premise of ensuring the peak signal-to-noise ratio of the image and the structural similarity,a brighter display effect is achieved.This paper verifies the algorithm using two virtual micro-lens implementations:software 3ds Max and CCD camera array.Experiments have proved that,both the virtualviewpoint rendering algorithm and the chroma compensation algorithm presented in this paper can achieve good results,and improve the subjective and objective quality of the virtual viewpoint images in order to enhance the color effect of stereoscopic display,whether for virtual or real multi-view systems.