The Research On Halftone Image Feature Extraction And Classification

Inverse halftoning technology will restore a halftone image into a continuous gray image.It is the opposite halftone process technology,which is widely used in image sharpening,resizing,color correction,image compression processing.Existing inverse halftoning divided into general inverse halftoning and special inverse halftoning.General inverse halftoning technique used to reconstruct some of the halftone image,since the lack of relevant halftone pattern information is difficult to obtain optimal reconstruction halftone image.Special inverse halftoning technology using specific halftone pattern characteristics and knowledge,in theory,get the best reconstruction particular halftone pattern,but in practice,since the image halftone mode is unknown,it is difficult to select some special type of inverse halftoning technology to achieve optimal reconstruction digital halftone patterns.However,special inverse halftoning technology problems can be solved by identifying the halftone image of halftone patterns.So get a halftone image classification method is very important before the inverse halftoning.This paper based on the relevant halftone image classification algorithm made a more in-depth research.The main work is as follows:(1)This paper combined with local binary pattern of thought,put forward category of neighbor pixel correlation descriptor.The algorithm first acquires neighboring pixels correlation diagram.Second,consider the three factors of correlation between the pixels,the distance between pixels and the angle between pixels then design feature extraction method of pixels correlation.Finally,classify experiments to verify the effectiveness of the method using BP neural network.Experiments show that neighbor pixel correlation descriptor algorithm than the existing method for error diffusion classes have improved in terms of performance computational complexity,recognition accuracy.(2)Covariance matrix in halftone image feature extraction for the problem of longer time and lower correct classification accuracy,this paper presents a classification method based on improved covariancematrix in the halftone image.According to the covariance matrix in achieving halftone image classification number is small and does not reflect the characteristics of their local and global information,then to improve the underlying characteristics of the covariance matrix.Firstly,the horizontal edges operator and vertical edges operator by the operation after generated two values,then these two features added to the underlying characteristics of the covariance matrix for the halftone image feature extraction,and finally the image classification.Experiments show that the improved covariance matrix method can improve the correct classification rate of the image,and the classification of the more stable performance.