Applied Research Of Image Grouping Technique In Document Image

With the development of computer and image data acquisition technique, digital document image apprear more and more, in the same time, the character recognition technology and document image processing technique are gradually maturing. Office automation based on document image is paid attention to researcher, applying also more and more extensive.In this paper, we will study and analysis the problem of definition for document image.Definition of image is an subjective measurement of human vision.Evaluation function mostly based on image histogram, variance of energy, square gradient of energy , information entropy,frequency spectrum, Laplacian energy, variance of gauss, square gauss gradient and Laplacian-Gauss.These method can evaluate the definition of image which caused by the defocus.Texture of document image is prominence, study on definition for these images is few.From the entropy of information ,the entropy of an blur image may be high,may be low, so these methods is not suitable for document images.This paper makes a study on how to evaluate definition of document images. Based on the texture of document images, we propose an algorithm based on the image entropy and histgram, also we defined the function of the document image definition. This algorithm is tested through trials. The experiment results indicate this algorithm is robust and effective.Defination evaluation of document images can eliminate the influent which caused by different person, gurentee the consistent of evaluation.During the cosmically document image prossing, auto-evaluate system garentee the effective and quantity, reduce the intensity of human.The main achievements of this thesis are as following: 1.Analysis the texture of document image in space domain and DCT domain.2. Give an evaluation model for document image definition based on entropy and pixel distribution3. The experiment results indicate this algorithm is effective.4. Experiment result analysis and future prospect . Analyse and summarize current work and make a outlook to the forward work.