Research On Image Recovery And Anti-noise Performance Improvement Algorithm In Bayer Pattern

For the sake of economical practicability and portability,most commonly used digital camera devices use a single sensor for image acquisition,which makes the obtained image only one color channel of red,green and blue at each pixel position.The pixel value.The process of reconstructing missing pixel values by algorithm processing to obtain a full color image is called demosaicking.Demosaicking is the core technology of digital cameras.It has a significant impact on image denoising,super-resolution,image recognition and other subsequent operations,and has significant research significance and commercial value.At present,the effect of each demosaicking mosaic on image edge reconstruction still has certain pseudo-color,zipper effect,etc.,and most of them are reconstructed for color filter array without noise interference.Therefore,demosaicking algorithm with certain anti-noise performance has higher use value.This paper introduces the basic theoretical knowledge of color image demosaicking and the classic demosaicking algorithm,and proposes two methods of demosaicking for some shortcomings of the current algorithm.The main work of this paper includes:1.Bayer pattern image reconstruction algorithm for improving local correlation is proposed.The algorithm utilizes the correlation between natural images and the characteristics of guided filtering to perform local uniform verification on the improved block gradient calculation results when reconstructing the green channel,and nonlinear maps the verification results to direction weights,thereby obtaining more Accurate weight.The directional weight and the direction candidate interpolation are adaptively fused to reconstruct the green channel,and the reconstruction of the red and blue channels is further modified according to the reconstructed green channel,thereby obtaining a higher quality color image.Experimental comparisons in subjective and objective aspects show that the image reconstructed by this method has a clearer edge texture in detail.2.Reconstructed image quality and noise immunity improvement algorithm based on residual network.The algorithm equalizes the error pixel value optimization problem reconstructed by interpolation method to the denoising problem,and uses the advantage of convolutional neural network in denoising to inaccurate or erroneous pixel values in the green channel image reconstructed by block gradient adaptive interpolation.Optimize processing.The reconstructed green channel image then directs the initial reconstruction of the other two channels and further refines the reconstruction results to obtain a higher quality color image.The algorithm combines the advantages of convolutional neural network in denoising with the interpolation method.The results of the final data simulation show that not only the interpolation results are optimized,but also the anti-noise of demosaicking on Bayer arrays with noise interference.