A unified framework of deep unfolding for compressed color imaging

Abstract

Traditional iterative-based reconstruction algorithms for compressed color imaging often suffer from long reconstruction time and low reconstruction accuracy at extreme low-rate subsampling. This paper proposes a model-driven deep learning framework for compressed color imaging. In the training step, extract the image blocks at the same position of the R, G, and B channel images as the ground truth, then singular value decomposition is performed on the measurement matrix to obtain the optimized measurement matrix and low-dimensional measurements, afterwards the ground-truth and optimized measurements are utilized to construct a large amount of training data pairs to train an 'end-to-end' deep unfolding model for compressed color imaging. In the test step, the single pretrained model is used to reconstruct high-quality images from optimized low-dimensional compressed measurements for each channel and synthesize a color image. Numerical experiments demonstrate that our proposed unified framework can achieve high accuracy and real-time reconstruction for the color image at extremely low subsampling rate.

Ultra‐High Detectivity MBene/Si vdW Schottky Photodiode for Color Single‐Pixel Imaging

“…The reconstruction of the color image involves selecting the image blocks located at the same position (x,y) from the three-channel images. The grayscale values (P c (x,y), c = R, G, B) of each pixel were retrieved and afterwards utilized in real-time to synthesize a color image with dimensions of 256 × 256 pixels, employing a specific technique, whose formula is as follows [43] ⎡…”

“…The reconstruction of the color image involves selecting the image blocks located at the same position ( x , y ) from the three‐channel images. The grayscale values (P c (x,y), c = R, G, B) of each pixel were retrieved and afterwards utilized in real‐time to synthesize a color image with dimensions of 256 × 256 pixels, employing a specific technique, whose formula is as follows [ 43 ] $$$$\begin{equation}\left[ { \def\eqcellsep{&}\begin{array}{@{}*{1}{c}@{}} {R\left( {x,y} \right)}\\ {G\left( {x,y} \right)}\\ {B\left( {x,y} \right)} \end{array} } \right] = \left[ { \def\eqcellsep{&}\begin{array}{@{}*{1}{c}@{}} {{P}_R\left( {x,y} \right)}\\ {{P}_G\left( {x,y} \right)}\\ {{P}_B\left( {x,y} \right)} \end{array} } \right],0 \le x \le 256,0 \le y \le 256\end{equation}$$$$Furthermore, the pixels of the reconstructed image are significantly influenced by the frame number of the encoded patterns ( F n ). This frame number is determined using a differential computation approach, as described below: $$$$\begin{equation}{F}_n = \ 2\ \times M \times N \times SR\end{equation}$$$$where M × N is the pixels of a reconstruction image, and SR is the sampling rate.…”