Font Size: a A A

Research And Design Of Lensless Camera Based On Compressive Sensing

Posted on:2020-10-11Degree:MasterType:Thesis
Country:ChinaCandidate:Y S ZhangFull Text:PDF
GTID:2392330620962279Subject:Information and Communication Engineering
Abstract/Summary:PDF Full Text Request
The price of traditional camera is expensive for the existence of the lens,Also,the lens lead to the volume of camera can not be reduced,which limits the use of the camera.In order to improve this situation,different kinds of lensless cameras was created,but there are still some shortcomings such as high cost,large overall size of the camera,low resolution of reconstructed images and long reconstruction time.Aiming to solve the above problems,this paper designs a camera based on coded aperture and compression sensing theory,cancels the camera lens through mask and reconstruction algorithm.While ensuring the image quality,the imaging time is reduced,too.At the same time,the low-cost mask material is used to integrate the camera photosensitive sensor to reduce the overall volume of the camera.The main research work of this paper is as follows:(1)For cancelling the existing camera lens,combining with the theory of compression sensing which can restore high-resolution images and the mask that can realize the measurement matrix physically,designing a mask before the photosensitive sensor,The camera imaging mode of the reconstruction algorithm using the compression sensing theory achieves the purpose of replacing the lens by calculation.(2)Aiming at the problem that the resolution of the existing lensless camera is low,the imaging time is long and the light throughput is insufficient,a mask based on the m-sequence generation matrix as a mask pattern has been designed.By verifying the incoherence of the m-sequence generator matrix and making it qualified as a compressed sensing measurement matrix,the binary matrix value balance of the m-sequence generation matrix is set to ensure that the ratio of the area of the light-transmitting area to the non-light-transmitting area is 1:1,to meet the highest threshold of 50% light throughput.The coding matrix of the mask pattern can be linearly encoded and projected onto the camera sensor to achieve the purpose of matching the subsequent image reconstruction algorithm,improving the operating efficiency of the algorithm and increasing the optical throughput to ensure the imaging quality.(3)In view of the existing camera's lack of bulk and high cost,the physical realization of the mask is set to chrome-plated glass to reduce the production cost,and the mask is directly attached to the camera.The distance between the sensor and the mask is only determined by the thickness of the mask,which directly reduces the volume of the entire camera.(4)Aiming to solve the problem that the camera image reconstruction algorithm run too long.Based on the idea of classifying images,a reconstruction algorithm based on the combination of total variation method and separation matrix method is proposed.In the case where the original image pixel is low,the algorithm processing speed is improved by performing singular value decomposition on the mask measurement matrix.In the case that the original image is a super-resolution image,the total variation method is used as the reconstruction algorithm and the block processing mode is added to improve the operation efficiency of the algorithm.This paper builds a complete experimental platform for a lensless camera.Design the optical path modulation module,data collection module and image reconstruction module,and analyze the experimental data in detail.The experimental results show that the mask based on the m-sequence generation matrix design can record the image that can be used for the subsequent reconstruction algorithm when the distance is 0.3-0.5m from the shooting source,and the image reconstructed by the total variation method under the under-determined linear programming condition.When the measurement group reached 500 times,the reconstruction rate did not increase significantly,and it could reach 80%.Under the condition of nonunderdetermined linear programming,the reconstructed image is reconstructed by the separation matrix method.When the number of measurements reaches 200 times,the reconstruction rate is not significantly improved,reaching 90%.
Keywords/Search Tags:lensless camera, coding mask, compression sensing, total variation method
PDF Full Text Request
Related items