Reconstructing of Images Application in Computer Vision and AI

Reconstructing image is a pivotal and crucial process in computer vision, powered by artificial intelligence. It entails the transformation of fragmentary, degraded, or low-resolution images into complete, enhanced, or high-resolution versions. This process plays a crucial role in various fields, including self-driving, restoration of historical heritage, satellite and aerial image processing and astronomy and space exploration, which will be mentioned and detailed in this paper. Reconstructing of images, a highly-developed technology, has already shown great effect in many fields such as medicine, searching for relics, and the main topic of this paper, computer vision. As the main part of the reconstructing of images, the algorithms of the probability density function of the measurements from the outset and derivation of a reconstruction scheme which is optimal in the maximum likelihood sense are really indispensable. Therefore, a method which can better reconstruct image and return a more unbiased images with sharper pixels can be found. The algorithm combines Poisson distribution and Bernoulli distribution with the maximum likelihood estimator. It is more advantageous compare to most of other methods: for instance, compare to Bayesian algorithm, which based on many certain but impractical probabilistic model assumptions, the method. It can return a more unbiased image with much fewer prior information. Thus, this method is quite suitable for the computer vision and AI identification, which usually need to identify the information of a blurry image. This paper is aimed to claim about how our algorithm worked and then explain why it is so suitable for AI identification to the blurry images.