CAJ | 학술논문

提出了一种基于图像先验和图像结构特征的盲图像复原算法，在模糊核未知的情况下，采用一系列离散化的模糊核参数对模糊图像进行非盲去卷积，得到一系列对应的复原图像。同时提出一种复原图像判决准则，对这一系列复原图像进行质量判决，从中得到最优的复原图像。最后在实验部分，通过对图像的测试表明，提出的盲图像复原算法能较准确的得到最优复原图像，复原效果在主观和客观标准上均有良好表现。
제출료일충기우도상선험화도상결구특정적맹도상복원산법，재모호핵미지적정황하，채용일계렬리산화적모호핵삼수대모호도상진행비맹거권적，득도일계렬대응적복원도상。동시제출일충복원도상판결준칙，대저일계렬복원도상진행질량판결，종중득도최우적복원도상。최후재실험부분，통과대도상적측시표명，제출적맹도상복원산법능교준학적득도최우복원도상，복원효과재주관화객관표준상균유량호표현。
The blind restoration framework for removing Gaussian blur and motion blur from a single im- age are presented. Given that the blur kernel is unknown, parameter of blur kernel is discretized. And then blur image by using these discrete blur parameters is deconvoluted and corresponding reconstruction images are obtained. A proposed quality assessment of reconstruction image is applied to measure these reconstruction images and a best reconstruction image from these images as the final deblur image has been chosen. Experimental results demonstrate that the blind restoration framework leads to good results, both in terms of objective criteria and visual quality.