Based On Image Recognition Painting Of Authenticity Identification Research

Oil painting as an important type in the western traditional painting art has a long history of several hundred years. It has its unique artistic language charm and unique traditional techniques. But due to various reasons, in the traditional paintings exists a lot of fakes, and regular, rely solely on expert empirical appraisal way has a biggest problem is that subjectivity is too strong, for some controversial works it`s hard to have a convincing objectively identify evidence, therefore, we need to explore a more effective, scientific, quantifiable paintings identification methods. This article through to the oil painting preprocessing, feature extraction and classification identification to get the authenticity of the probability. From two aspects of feature extraction algorithm and the design of the classification method, based on the analysis of comparative advantages and disadvantages of existing methods, put forward the improvement of feature extraction and classification algorithms.Local area of painting and calligraphy are studied. The oil painting art stylenfeature extraction method based on shape feature. In view of the traditional edgedetection algorithm on the edge of the ihs space is lost, that part of the edgediscontinuity problems, using a multi-scale based on ihs space structure color morphological edge detection algorithm. First defines the basic color morphology operator in HIS space, on the basis of using the different scales of structural elements to extract color image edges, and then using multi-scale merge algorithm to synthesize each edge to edge detection results are obtained. After using fixed module partitionmethod, is the most can reflect the characteristics of the wind local area; Finally on the basis of the calculation used to draw painting strokes morphological characteristics,manifests the draw style of painting. To the integral style of the paintings were studied.This paper introduces the global feature extraction method based on color and texture.The color feature extraction of its color histogram, texture feature extracting its gray level co-occurrence matrix, the characterization of the statistical features of the whole picture. Experimental analysis shows that the histogram can well reflect the differentartists on the whole space, screen layout, the imposing manner of artistic style.For any kind of characteristic can well express the problem of image content,proposed a based on feature level data fusion and decision level data fusion recognition framework combining. Research results show that using principal component analysis to feature level data fusion, can make the content, the characteristics of good expression the fused images at the same time also has reached the feature dimension reduction, get rid of the characteristics of redundancy between purposes; Majority vote on decision level data fusion is proposed, which can make full use of the complementarity between different classifiers, has achieved higher recognition rate than single classifier.