CAJ | 학술논문

提出了一种滤波窗口方位、尺寸和形状都可随图像纹理结构自适应变化的开关滤波算法。滤波窗的尺寸可随图像梯度值的变化而变化，滤波窗方位排列与图像对应点像素最小梯度值方向一致。同时，为了将受噪声污染的点和图像的细节纹理像素点分开，给出了一种噪声点检测判据和开关算法流程。与传统均值、中值及相关文献提出的自适应平滑滤波算法相比，由于本文算法在降噪的同时兼顾了图像的局部纹理分布结构，因此在保护图像细节方面做得更好。仿真实验结果证明了本文算法的有效性。
제출료일충려파창구방위、척촌화형상도가수도상문리결구자괄응변화적개관려파산법。려파창적척촌가수도상제도치적변화이변화，려파창방위배렬여도상대응점상소최소제도치방향일치。동시，위료장수조성오염적점화도상적세절문리상소점분개，급출료일충조성점검측판거화개관산법류정。여전통균치、중치급상관문헌제출적자괄응평활려파산법상비，유우본문산법재강조적동시겸고료도상적국부문리분포결구，인차재보호도상세절방면주득경호。방진실험결과증명료본문산법적유효성。
A novel on-off image filter algorithm is introduced in this paper, whose shape, size and orientation can vary with the image's regional structure. The filter window's size vary with the corresponding pixel's gradient magnitude, and it's shape, orientation is adjusted in such a way as to lie in the direction of the least gradient. Meanwhile, a criterion and flowchart which can classify pixels into noise-pixels and structure pixels is also proposed. Compare with the traditional mean filter, the traditional median filter and the method of an adaptive smooth filter algorithm of still images, the simulation results prove that the proposed adaptive window image filter algorithm have a better resolution and it can protects image's structure more well.