基于特征标记的多尺度建筑空间图像分解仿真

Simulation of Multi Scale Building Space Image Decomposition Based on Feature Mark

多尺度建筑空间图像分解重组过程中易出现阶梯现象,导致图像显得轮廓过分尖锐,影响分割结果,提出一种基于特征标记的图像分解方法。通过主流色标记符与颜色空间标记符量化标记颜色方差与图像空间结果,标记图像纹理和轮廓特征,根据标记结果构建分解模型,克服全变差重组图像时引起的阶梯现象;将像素灰度值对应坐标垂直投射至坐标平面内,挑选局部极值点、三角剖分与三角差值曲面处理,构建差值曲面当做包络上界,局部极小值迭代至出现包络下界,设定极大值与极小值包络曲面的代数均值为图像分解阈值,对比像素灰度值坐标完成多尺度建筑空间图像分解。实验结果表明,图像的像素分解与纹理分解结果较为准确。

During the decomposition and reorganization of multi-scale architectural spatial images,sharp image contours are caused by the ladder phenomenon,affecting the segmentation results.Therefore,in this paper,an image decomposition method based on feature markers is studied.Firstly,according to the mainstream color markers and color space markers,the color variance and image space were quantified and marked,and the image texture and contour features were also marked.Secondly,the decomposition model was founded by labeling results in order to overcome the ladder phenomenon caused by total variation image reconstruction.Then,the coordinates corresponding to the pixel gray value were vertically projected into the coordinate plane to select local extreme points,triangulation and trigonometric difference surface.Then,the difference surface was established to act as the upper bound of the envelope,and the local minimum was iterated until the lower bound of the envelope appears.Then,the algebraic mean of the maximum and minimum envelope surfaces was set as the decomposition threshold of the image.Finally,the pixel gray value coordinates were compared to achieve the image decomposition of multi-scale building space.The experimental results show that this method has accurate image pixel decomposition and texture decomposition.