Adaptive image restoration method and device utilizing gradient properties

• Main Authors: XIONG RUIQIN, LIU HANGFAN, GAO WEN, MA SIWEI
• Format: Patent
• Language: English
• Published: 03.06.2015
• Subjects: CALCULATING; COMPUTING; COUNTING; IMAGE DATA PROCESSING OR GENERATION, IN GENERAL; PHYSICS

The embodiment of the invention provides an adaptive image restoration method and device utilizing gradient properties. The method mainly comprises the following steps: building training samples of the gradient data of to-be-restored pixel positions in an image; calculating the statistics properties of the gradient data of the to-be-restored pixel positions according to the training samples; carrying out constrained optimization on the image according to the statistics properties of the gradient data of the multiple to-be-restored pixel positions so as to obtain reconstructed pixels corresponding to the to-be-restored pixel positions in the image. According to the embodiment of the invention, the gradient data of non-locally similar image blocks are used as the training samples of the to-be-restored pixel positions in the image, the statistics properties, such as expectation and variance, of the gradient data of the to-be-restored pixel positions are calculated according to the training samples, reconstructed pixel positions corresponding to the to-be-restored pixel positions are obtained, and one novel adaptive image restoration method based on gradient statistics properties is provided. The method can adapt to the condition that different pixel positions in the natural image correspond to the statistics properties of different gradient data.