The Research And Modeling Of Imaging Process In Image Super-Resolution

High-resolution images play an important role in modern life and applications. Various industries and fields have a strong demand for high-resolution images. As access to high-resolution images from the means of the hardware method is limited by the cost, technological level, so the use of a signal processing method of obtaining a high resolution image from a low resolution image sequence is necessary, which is called super-resolution images. Super-resolution is a necessary element in many applications, and is now the hot topic of research in image field. In the field of super-resolution image, the research and modeling of imaging process has been difficult, which is an important factor in whether the super-resolution technology can have a forward propulsion. But there are some obvious drawbacks on the research of super-resolution imaging model, namely the use of a simple mathematical model, without considering the impact of PSF. It does not reflect the true and accurate imaging process.The paper study in-depth the imaging process of real-world scenarios to low-resolution images, analyzes the important factors that lead to image degradation, propose the modeling methods to accurately estimate the point spread function based on the imaging process. This paper estimates the point spread function by the fuzzy edges precise positioning and precise edge definition estimated non-blind method. This method has high accuracy of edge positioning prediction and calculation of the point spread function.In order to demonstrate the accuracy and the superiority of the proposed imaging model, the paper introduced the imaging model into the traditional image super-resolution algorithm called maximum a posteriori (MAP) algorithm for high-definition image reconstruction, and elaborated the algorithm derivation. Advantage of this method is demonstrated by comparing the super-resolution reconstruction result at last.In the paper, the point spread function estimation algorithm consists of three parts, in the first, we make the preprocessing of a particular input calibration board image, determining the location of the blur edge by corner location and curve fitting, then, we get the maxima and minima in the normal direction of fuzzy edges with bilinear interpolation and RANSAC method, then estimated the position of high-resolution edges in sub-pixel. Finally, calculate the point spread function by the least squares method and Newton fastest gradient descent method.Finally, the paper proposed the assessment methods and standards to evaluate the PSF and imaging model and analyze the reason of the error produced. The paper also compared with the international forefront of several algorithms, and made a summary of the accuracy of the algorithm in robustness and the calculation speed.Experimental results of this paper prove the point spread function estimation algorithm proposed in this paper can accurately portray the image degradation process, comparing with the traditional general imaging model, the imaging model including the point spread function to reflect the imaging process of real scene more accurately, which effectively improve the effect of super-resolution image reconstruction.