On Compressed Color Imaging Via Random Coded Aperture

Compressed sensing is an innovative theory of signal acquisition and processing. This theory states:the sparse or compressible signal can be sampled with a rate much lower than the Nyquist sampling frequency, and the original signal still can be accurately reconstructed by using the appropriate recovery algorithms. After CS theory was proposed, it has attracted many attentions, with the development recent years, which have been applied in many areas of signal processing, especially for imaging application.Compressed imaging can be divided two stages:coding (sensing) step and decoding (reconstruction) step. The main idea is to extract high-quality images from undersampled date through effective calculation methods. One typical example of compressed imaging is the single pixel camera (SPC) developed by Rice University, however, the most important disadvantage for SPC is that it can only capture one measurement with single shot. Compressed coded aperture imaging can overcome this disadvantage effectively, and it can capture all measurements with single shot. In Compressed coded aperture imaging system, a coded aperture is added in optical imaging system, and a more complex point spread function (PSF) than the pinhole aperture is introduced, which armed to increase the light hitting a detector in the-optical system without sacrificing resolution and coded image; in reconstruction step, it restores the high quality original image from coded measurements with the aid of compressed sensing reconstruction algorithms, and realize the compressed image reconstruction. The experimental results show that under the conditions of same size and resolution, compressed coded aperture imaging method can be used to improve the light utilization efficiency of focal pixel arrays, and the random modulation property of coded aperture mask can be used to record more information, a higher quality image is reconstructed, which have good robustness against noise.By the work on compressed reconstruction and the super-resolution imaging of the gray images, the thesis extends compressed coding aperture imaging to color imaging application and several issues around compressed coded aperture color imaging theory, reconstruction algorithms and optical imaging system.The main work and innovation of this thesis are summarized as follows:(1) The basic content of compressed sensing, the relevant knowaledge of the traditional pinhole camera imaging and coding aperture imaging are reviewed. On this point, the main purpose of introduction of the compressed coding aperture imaging is analyzed, the compressed coding aperture imaging (CCAI) methods from the perspective of measurement matrix structure and imaging system is studied. On the standard gray images, the numerical experiments uses block compressed sensing method to compare the imaging quality and noise robustness with the coded aperture or not at different sampling rate to prove the effectiveness of the compressed coding aperture imaging method.(2) Compressed coded aperture imaging is extended to the color imaging area, and a color imaging system is proposed:the compressed coded aperture color imaging system based on the RGB channels (CCACI). CCACI system uses different coded apertures to implement compressed measurement and reconstruction for RGB channels independently, then the final color image is synthesized. On the standard color images, the CCACI method compares the reconstruction quality and noise robustness at different sampling rate and proves the effectiveness of the compressed coded aperture color imaging system based on the RGB channels.(3) Due to the high correlation in RGB channels of color image. Taking advantage of the correlation of the RGB channels, a novel color imaging system is proposed:the correlation-channels-based compressed coded aperture color imaging (CCCACI), the correlation channels of color images is taken into the imaging system, the same coded aperture is used to capture the compressed measurement of the RGB channels with single shot, and then the color image is reconstructed. On the standard color images, the effectiveness of CCCACI is proved by comparing the reconstruction quality and noise robustness results at different sampling rate.