Research On Depth Image Quality Enhancement Algorithm

The depth image is an image that represents the distance between the depth sensor and the scene object.With the rapid development of depth sensors,depth images have been widely used in many fields such as human-computer interaction,3D video,and virtual reality.However,due to the limitations of the acquisition equipment and the influence of the acquisition environment,the directly acquired depth images have problems such as depth data loss and low resolution,which seriously affect the subsequent application of the depth images.According to the problems in depth images,this paper proposes corresponding restoration and upsampling reconstruction algorithms.Research on the repair of the hole formed by the loss of data in the depth image,a depth image repair algorithm based on Curvature Driven Diffusion and edge reconstruction is proposed to achieve the repair and completion of a single depth image;address the problem of low depth image resolution,A depth image upsampling reconstruction algorithm based on global edge model is proposed,which realizes the upsampling reconstruction of a single depth image and obtains a higher resolution depth image.The main research work and results are as follows:(1)The lack of depth data leads to a large number of holes in the depth image.The use of traditional repair methods can easily cause the edge of the image target object to be over-filled or under-filled,causing problems such as edge distortion and boundary blur.To solve this problem,this paper proposes a depth image restoration algorithm based on Curvature Driven Diffusion and edge reconstruction.The algorithm uses a Curvature Driven Diffusion model,and uses effective pixel information in the image neighborhood to diffuse the local structure from the outside of the holes to the inside,which can accurately fill the holes.Then,a binary segmentation filter is designed to perform edge segmentation on the fuzzy area of the filled pixel,and the Markov Random Field model is used to reconstruct the texture of the blurred edge.The binary segmentation filter can effectively improve the accuracy of the Markov Random Field for edge reconstruction,degree.The experimental results show that the proposed algorithm can effectively repair holes and eliminate the blurring of object edges.Compared with the JBF,FMM and CDD repair algorithms,the repair effect of the holes at the edges of the objects in the depth image is greatly improved,and the average gradient index result also has obvious advantages.(2)The resolution of the depth image is low,and the image after upsampling and reconstruction is prone to ringing,artifacts and blurred edges.To solve this problem,this paper proposes a depth image upsampling reconstruction algorithm based on global edge model.The algorithm combines interpolation and optimization methods.First,the low-resolution image is reconstructed by interpolation;then,the reconstructed image is globally optimized.The regular term design of the optimization function uses the combination of Tikhonov operator and the global edge model.Finally,the constrained optimization problem is solved by the multiplicative alternating direction algorithm to obtain high-resolution images.The Tikhonov regularization operator can smooth the image and suppress noise well;the global edge model can provide a continuous domain representation of the image edge in the local area and the entire image,and provide edge prior knowledge for optimization.The algorithm applies the optimization function to the interpolated image because the interpolated depth image can obtain more edge pixel information and obtain a more accurate edge model when estimating the global edge model.The experimental results show that the up-sampling reconstruction algorithm proposed in this paper can effectively improve the reconstruction effect of low-resolution depth images.Compared with Bicubic interpolation,SC,and the current deep learning-based SRCNN and FSRCNN algorithms,the PSNR and SSIM evaluation indicators have been significantly improved,and can effectively alleviate the problems of ringing,artifacts and blurred edges.