Research On Score Based Synthesis Of Structured Decoration Picture

The artistic expression form of decoration picture belongs to painting. Compared with realism painting, the decoration picture strongly emphasizes freehand brushwork and does not rigidly adhere to form. The decoration picture making is not a stick to fine brushwork confined to theme moreover it does not lay much stress on perspectivity and structure relation. It embodies internal word of artist better than the realism painting. The structured decoration picture, one kind of common decoration picture, usually appears on book inset, cards making, china designing, and frame decoration. Some expression characteristics of details on structured decoration picture are fine line and complicated texture. It has some self-similar natures. The themes of structured decoration picture are mainly spray, branches, leaves and other patterns from the nature. But it is hard to drawing the structured decoration picture, so we research on structured decoration picture and put forward a method for synthesis of structured decoration picture.This thesis parameterizes the sampling elements in the real structured decoration picture, collects the characters of sampling elements and then synthesizes a brush of new decoration picture using the sectional matching deformation and other graphics methods.The style of sampling elements decides the style of synthesized decoration picture. The style and texture characteristics of original picture can be well saved if the sampling elements are collected from the real parameterized decoration picture. This thesis applies closed image matting for colleting sampling elements from the real decoration picture. The u and v parameters of sampling elements (upper profile, edge of regression and below profile) are collected after section process, and then distance, width and curvature, three characteristics of each section are calculated.During the synthesis of the new painting brush of structured decoration picture, the u and v parameters of input brush are parameterized in the same method. Firstly, determine the section of painting brush awaiting matching in the samples with the method of dynamic programming. Use the deformation method based on the equilateral triangle to deform the sections which are matched and make its profile and edge of regression coincide with the input profile and edge of regression. Secondly, cut the overlapped parts which are matched in the adjoining section using the image segmentation algorithm with the minimal cutting costs and at the same time make the profile as smooth as possible. Finally, use method of texture synthesis to remedy the defects of serration, distortion and etc. caused by image segmentation and synthesize one stroke of new structured decoration picture. The new structured decoration picture can be painted after repeating the aforementioned process. Finally, the synthesis result can keep the original sample style and result was good.