Research On Cosinusoidal Encoding Multiplexed Multispectral Ghost Imaging

Optical imaging technology plays a vital role in modern life.People use optical imaging technology to perceive the world and record their life.Compared to the original black and white fixer camera(which uses the single band imaging technology),the trending "megapixel" level color camera(which uses the multispectral imaging system)can obtain richer information of the target scene.Moreover,multispectral imaging technology can improve accuracy and reliability in food safety,medical diagnosis,material analysis and target tracking and recognition.Therefore,exploring the more advanced multispectral imaging technology can better fulfill people’s pursuit of a high-quality life and the inevitable requirement of improving modern national defense.In some exceptional cases(e.g.,low-light environments,non-visible bands,multi-dimensional imaging),traditional imaging technology which uses a planar array detector always provides poor or even failed imaging results.To overcome this,previous research proposed ghost imaging technology to break through the technical barriers of traditional imaging.Ghost imaging is a computational imaging technology that mainly generates space light with known light intensity distribution by encoding and modulating the spatial light field.After irradiating the target scene,a single pixel detector is used at the receiver to collect transmitted or scattered intensity signals,then recovering the object’s two-dimensional spatial information according to the intensity correlation characteristics of the light field.Joint information reuse technology can achieve high dimensional imaging(e.g.,3D,multispectral,polarization,multitarget imaging).In recent years,researchers have conducted in-depth research on multispectral ghost imaging,put forward many creative ideas,and achieved fruitful results.However,the existing multispectral ghost imaging technology still fall behind traditional imaging technology in terms of imaging quality and efficiency.In order to improve the imaging quality and information acquisition efficiency of multispectral ghost imaging technology,based on the previous research,this paper proposes new methods and carries out the corresponding theoretical analysis,numerical simulation and experimental research.The main research work of this paper is summarized as follows:(1)Aiming at the low information acquisition efficiency of the existing multispectral ghost imaging technology,in Chapter 3,we propose the Hadamard-based cosinusiodal encoding multiplexed multispectral ghost imaging technology.First,we construct three coding matrices of cosine structure with different spatial frequencies according to the requirements of spectral channels.Then we fuse the coding matrices with the orthogonal Hadamard patterns to generate the structured light to illuminate the target objects.Next,we use a single pixel detector to collect the backward reflection signals.In the image restoration,we apply a linear algorithm to recover the mixed spectral image of the object.Based on the Fourier frequency shift characteristics of the cosine coding matrices,the multi-channel spectral mixed information can be converted from the spatial domain to the Fourier frequency domain for spectrum recombination,and then the spectrum of each spectral component can be extracted by an ideal low-pass filter to realize the decoding and separation of the mixed information.Finally,the spectral component image can be obtained by inverse Fourier transform of the isolated single-channel spectrum,and then fused into the multispectral image of the target object.The rationality and effectiveness of the proposed approach are verified by theoretical analyses,numerical simulations and preliminary laboratory experiments.The Hadamard-based cosinusiodal encoding multiplexed multispectral ghost imaging technology uses the Fourier frequency shift characteristic of cosine coding matrix to encode and reuse multi-channel spectral information in projection illumination,and decodes it in Fourier frequency domain in restoration stage.Under the premise of image quality assurance,this research synchronously restores the multispectral image of the target object from a projection cycle,and applies a digital image processing method in image restoration,thus greatly reducing the computational time of reconstruction and improving the efficiency of multispectral imaging.(2)The Hadamard-based cosinusoidal encoding multiplexed multispectral ghost imaging uses an ideal low-pass filter to extract and separate the multi-channel spectral information,resulting in an evident ringing phenomenon in the restored image,which reduces the quality of multispectral imaging.Chapter 4 analyzes the proposed Hadamard-based cosinusoidal encoding multiplexed multispectral ghost imaging from the aspect of filter parameter selection to improve the quality of multispectral imaging.The effects of different combinations of low-pass filter parameters(filter type and filter radius)on multispectral image restoration results are compared and analyzed through numerical simulation and experimental verification.The results show that the Gaussian low-pass filter is more advantageous than the ideal and Butterworth filters,and higher quality multispectral images can be reconstructed with a smaller filter radius using the Gaussian low-pass filter.After obtaining a filtering strategy,the proposed technique is compared with the existing multispectral ghost imaging technology.The experiment and analysis prove that the proposed Hadamard-based cosine-based multiplexing multispectral ghost imaging technology is advanced in terms of image quality and reconstruction time.(3)In order to explore more possibilities of cosinusoidal encoding multiplexed strategy and further improve the imaging efficiency of multispectral ghost imaging technology,in Chapter 5,we propose a cosinusoidal encoding multiplexed multispectral ghost imaging technique based on discrete cosine transform(DCT),which can efficiently recover multispectral images of target scenes with sub-Nyquist sampling volume.The core idea is to reuse the cosine structure coding matrices with definite frequency and the orthogonal discrete cosine basis patterns into the color structured light pattern to modulate the illumination on the target object,and use a single-pixel detector to collect the backscattered signal.The collected signals correspond to discrete cosine transform spectrograms of the multi-channel mixing information.The energy of discrete cosine transformed natural signals is concentrated in the low-frequency region,yet cosine coding matrices of different frequencies will shift the channel information to high-frequency areas through spectrum frequency shift.In the restoration phase,the inverse discrete cosine transform(IDCT)is applied to the spectral coefficients of each large-weighted channel to recover the spectral components,thus reconstructing the high-quality multispectral image.Numerical simulations and preliminary experiments verify the effectiveness of the proposed method.The proposed method can efficiently obtain the multispectral information of the target scene at a low sampling rate and reduce the time of projection and reconstruction computation.The results also show that even if the sampling rate is as low as 5%,the multispectral information of the target scene can still be identified.