An Algorithm For Reconstruction Of The Image Captured From Screen

Screen image is the image produced by the sampling process of digital camera taking electronic screen.Nowadays,digital cameras have become the mainstream instead of film cameras,which are mounted on smart phones and become an inseparable part of modern people's daily life.On the other hand,various kinds of display screens such as LED,LCD,DLP can be seen everywhere in daily life.As the main electronic information output equipment,they have become one of the representative products of modern life,and have already involved all aspects of life.Therefore,it is one of the most convenient and efficient ways to capture and store digital information by directly shooting the screen of display screen with digital camera.In the process of image digitization,the problem of image distortion caused by the imbalance of sampling frequency and signal spectrum has a long history.The traditional Nyquist sampling theorem shows that when the sampling frequency is greater than or equal to the maximum frequency of the original signal,the complete information of the original signal can be restored by the sampled signal.The physical structure of digital camera and display screen and the internal image processing algorithm affect the image of the camera screen.The image distortion caused by the frequency spectrum and sampling frequency of the image signal is particularly serious.In order to solve the problem of image distortion caused by the misalignment of image signal spectrum and sampling frequency,and to eliminate the tilt of shooting angle and lens distortion of the screen image,this paper studies the signal and causes of the screen image.By combining the idea of content-sensitive non-equal scaling and super-resolution reconstruction,the idea and algorithm of image reconstruction for the screen are proposed.The algorithm flow is attached to realize it.Compared with the traditional interpolation and scaling algorithms,the reconstruction algorithm accurately restores the spatial structure and color channel of the screen image to the original level,thus effectively suppressing the problem of signal sampling and spatial distortion of the image.The main work of this paper is as follows:Selecting different spatial frequencies and offset angles as sample sets,the signal model of the camera image and the corresponding physical structure of camera sampling and screen output are studied in depth,and the image signal is analyzed in detail.Based on the prior knowledge of LCD dot matrix structure,the scheme and process of space reconstruction for video screen images are proposed and described in detail.A linear filtering algorithm with adaptive length is proposed to separate the high frequency signal of the camera image.For the pre-processed image,a fast Greedy segmentation algorithm based on domain search is used.The segmentation algorithm based on target geometry and statistical model and Bernoulli random walk segmentation algorithm are used to complete the segmentation of LCD points in the screen image.Finally,the mapping relationship between LCD point region and reconstructed image coordinate system is constructed by using spatial reconstruction algorithm.Sample experimental results show that the adaptive length linear filtering algorithm can effectively component the high-frequency image mesh signal,and the three iterative segmentation algorithms can effectively extract the high-frequency image signal and establish the signal model.In order to obtain accurate and effective color channel reconstruction algorithms for screen images,experiments are carried out on the average interpolation algorithm,the interpolation algorithm based on extremum solution,the interpolation algorithm based on denoising,the interpolation algorithm based on interpolation product and the interpolation algorithm based on mathematical modeling.At the same time,the interpolation algorithm is compared with the classical scaling algorithms such as nearest interpolation,bilinear interpolation,bi-cubic interpolation,Lanczos interpolation and the interpolation algorithm based on denoising.Resampling interpolation is the most effective method to suppress scaling Moire pattern.The most effective color interpolation algorithm is obtained by comparing the Moire pattern observation,gray histogram,histogram equalization,histogram similarity and Papanicolaou distance of the experimental samples.Sample experiments show that the inner product based color reconstruction algorithm and denoising based color reconstruction algorithm are superior to other reconstruction methods,and obviously superior to the traditional scaling algorithm.