Research On The Key Technologies Of CMOS Image Sensor Based On Image Super-Resolution Reconstruction

For high-speed image acquisition system applications, the patial resolution, the dynamic range and the signal to noise ratio are the key parameters. High quality image acquisition and super resolution reconstruction are valuable both in theoretical research and engineering domain. In this paper, the technologies of high speed, high resolution and high dynamic range are explored. The key issues of CMOS image sensor technology, image denoising and image super-resolution reconstruction are studied.(1)Based on total variation model algorithm and with the basis of P-M regularization model, a weighted variational image denoising algorithm is proposed with specially selected non-increasing diffusion function. The weighted variational denoise algorithm has effective effect on salt and pepper noise while performing image processing at the same time. The image edge information is protected effectively. Compared with the classical P-M regularization model, the image processing can get a better the peak signal to noise ratio and achieve a good denoising effect, while requiring a reduced number of iterations.(2)Based on the regularization method, the image prior model using the Lorentzian distribution function is introduced, and a new lorentzian-based super-resolution algorithm(LBSR) is established. Experimental results show that the super-resolution reconstruction of Maximum A Posteriori(MAP) algorithm can improve the stability of the super-resolution reconstruction algorithm using a Lorentzian distribution of prior image model. It also improves the performance of image reconstruction. The algorithm can effectively overcome the unstable shortcomings of some traditional algorithm and obtain better image reconstruction quality.(3)A regression algorithm of the Linear Prediction Weighted Average(LPWA) interpolation algorithm is presented to perform the Bayer image interpolation reconstruction in CMOS image sensor. The LPWA algorithm reduce the strength of image smoothing,and to the image edge and texture is well protected.According to the analysis of the experimental results, the LPWA algorithm has better image interpolation effect with high efficiency. With the reduction of the algorithms complexity and computational steps, the LPWA interpolation algorithm can improve the high speed capabilities of image processing system.(4)For high-speed, high resolution image acquisition system,the performance of CMOS image sensor technology has much importance. Two new structure of digital pixel sensor circuit are proposed. An asynchronous self reset pixel circuit structure is designed and optimized.The size of the transistors is minimized to meet requirements of process dimension and layout implementation which occupies the smallest layout area, and is conducive to improve the pixel fill factor. Based on image super resolution reconstruction of CMOS image sensor architecture, the LPWA interpolation algorithm and LBSR algorithm is integrated in the image sensor. As a result of the asynchronous reset digital pixel circuit technology, the synchronous exposure of full pixel plane through the synchronous shutter can be realized.The reading speed of the pixel data is improved with a column parallel readout method. With the LPWA hardware interpolation algorithm, the high-speed data acquisition, reconstruction and transfer of the color image can be realized.