Single Image Reconstruction Algorithm Based On Multiple Observations

Due to the limitations of natural conditions such as scene environment and lighting,as well as the deviation of the hardware device itself,the image quality captured by most image sensors through a single observation is often unsatisfactory,and it is difficult to fully present the scene.Therefore,more and more equipment manufacturers and related research have begun to focus on the method of obtaining high-quality images by observing the scene multiple times and obtaining multiple scene images.The premise of this method is to complete multiple observations of the scene through hardware devices,so can the process of multiple observations of the scene be simulated from the perspective of the algorithm?In view of this,a single image reconstruction algorithm based on multiple observations is proposed,which realizes the re-observation of the original image and simulates the process of multiple observations of the scene.The whole algorithm process does not introduce complex optimization calculations,but is only driven by the original image data itself,and finally obtains high-quality images that conform to subjective vision and present better scene details.The implementation idea of this algorithm is to first consider the factors affecting image imaging,take the image pixel value range and pixel value statistical distribution as the basis,design the observation algorithm,determine the observation area,and then use the observation activation function to process the image data information of the observation area,distinguish the observation area from the non-observation area,and generate multiple observation images.Next,with the help of the characteristics of the observation activation function,determine the weight function of the observed image in the process of image reconstruction,measure the credibility of the observed image data,give weights to different image data,and generate a pixel-by-pixel weight matrix of the observed image.In the stage of image reconstruction,refer to the strategy of multi-scale image fusion,construct Gaussian pyramid and Laplace pyramid on the weight matrix and image data respectively,and then calculate the weighted sum of all observed images and the corresponding weight matrix at each layer of the pyramid,and finally reconstruct the pyramid based on the pyramid framework,and the result image is reconstructed.The resulting images obtained by the algorithm in this paper perform better in terms of information entropy,spatial frequency and average gradient,and are also in line with human subjective vision.In the first and second chapters of this paper,the research background,current status and research-related algorithms are introduced.In the third chapter,the algorithm model proposed in this paper is introduced in detail and explained in detail.In the fourth chapter,specific experiments are introduced with example images,and control experiments are carried out with LLF algorithm and BIMEF algorithm.Combined with the subjective and objective evaluation,it can be proved that the algorithm in this paper has a good and effective processing effect for general images.The algorithm proposed in this paper is expected to be applied to imaging devices such as smartphones and digital cameras to obtain high-quality images,and can also be applied to image dehazing and dark light image enhancement.