CAJ | 학술논문

Mean Shift是目前为止特征空间分析的最好方法之一，但其分割结果受带宽参数的影响。图像粗糙度是与视觉感受相关的图像纹理特征，对图像纹理的描述能力很强。图像像素的平均偏移量也体现了图像像素的总体离散情况。通过对高斯核函数的创建以及图像粗糙度的描述，创新性地给出了Mean Shift的窗口尺寸选择方法以及图像像素平均偏移的计算，仿真结果表明，该算法对不同类型的图像，均能得到令人满意的效果。
Mean Shift시목전위지특정공간분석적최호방법지일，단기분할결과수대관삼수적영향。도상조조도시여시각감수상관적도상문리특정，대도상문리적묘술능력흔강。도상상소적평균편이량야체현료도상상소적총체리산정황。통과대고사핵함수적창건이급도상조조도적묘술，창신성지급출료Mean Shift적창구척촌선택방법이급도상상소평균편이적계산，방진결과표명，해산법대불동류형적도상，균능득도령인만의적효과。
Mean Shift is one of the best ways of analysis in the feature space, but the segmentation result is bound up with band-width. Image coarseness is the image feature which is related to visual perception and is good for image texture description. Image pixel Mean Shift reports the scatter of an image. This paper creatively proposes a Mean Shift window size selection approach and the Mean Shift of image pixel by the creation of Gaussian kernel and the description of image coarseness. The experimental results show that the algorithm can get satisfactory effect for different types of image.