Discrete Cosine Transform Filtered Correlation Imaging With Compressed Sensing

Correlation imaging is a new imaging scheme introduced in recent years.We know that in geometric optics,object point and image point in the optical path maintain a one-to-one corresponding imaging relationship.In correlation imaging,the imaging system is divided into two spectral path,one is the reference end,the light path does not contain any object information,by a surface diagnosis detector used for imaging to detect the signal;The other part is the probe end,which only contains the information of the object in the optical path,but only by a bucket detector without imaging function to collect the signal.We were surprised to find that neither the reference end signal nor the probe end signal alone could recover the information of the object.However,the image of the object can be reconstructed by associating the two signals.Therefore,compared with geometric optical imaging method,correlation imaging method has many unique advantages :(1)it can realize non-local image reconstruction;(2)Lens-free imaging can be achieved;(3)During the imaging process,it has a strong anti-interference ability and can resist atmospheric turbulence and the impact of scattering media on the imaging quality;(4)Can realize single pixel imaging,etc.Although associated imaging technology has many advantages mentioned above,it also encounters many difficulties and is limited by many factors in practical application.First,when conducting associated imaging experiments using thermal light,it is found that the imaging resolution is not high,the peak signal-to-noise ratio of reconstructed images is low,and the imaging quality is not ideal.Secondly,a large amount of sampling data is needed to reconstruct the image using the traditional association imaging scheme.As a result,it takes up a lot of storage space and takes a lot of time to complete the calculation.Therefore,how to solve these problems has become the focus of scientific researchers,is also the focus of this paper.In order to solve the problem of low resolution of reconstructed image,we take advantage of the feature that discrete cosine transform function can highly concentrate the image information,realize the transformation of image from spatial domain to frequency domain,and perform threshold filtering operation on the image frequency domain information,and propose an associated imaging scheme based on discrete cosine transform filtering.Compared with the traditional relational imaging scheme,the resolution of reconstructed image can be increased to more than 10 times when appropriate threshold value is selected,and the target object is clearly recognized.At the same time,the scheme also greatly reduces the number of samples required for image reconstruction.With a small number of samples,most of the information of the original object to be measured can be reconstructed clearly,which greatly saves the calculation time of reconstruction.However,through comparative analysis of reconstruction results,it is found that the associated imaging method based on discrete cosine transform filtering is very sensitive to noise,and certain background noise will be generated during image reconstruction,affecting the quality of reconstructed images.In order to solve this problem,we propose a compressed correlation imaging scheme based on discrete cosine transform filtering combined with compressed sensing theory.Experimental results show that this imaging scheme can eliminate the influence of background noise.And,combined with traditional association and compression perception theory solution compared to the compression associated imaging scheme based on discrete cosine transform filtering can effectively improve the peak signal to noise ratio of reconstructed image,reduce the reconstructed image and the original error between the object under test image pixels,improve the reconstructed image and the original of the object under test image similarity degree,improve the quality of the reconstructed image.More surprisingly,the introduction of compressed sensing theory can further reduce the number of samples required for image reconstruction,and further shorten the computing time required for image reconstruction.In addition,this paper also uses this scheme to explore the relationship between selecting different thresholds and reconstructed image quality.The results show that under the same number of samples,with the decrease of threshold value,the quality of reconstructed image is constantly improving.