CAJ | 학술논문

提出了一种针对脉冲噪声图像的边缘检测算法，算法借鉴了中值滤波的思想，并采用十字型卷积模板计算图像梯度。首先，对参与图像中梯度计算的像素点进行阈值判断，如果是噪声点，该点像素值用3x3窗口中值滤波结果值替代，然后参与梯度计算，如果不是噪声点则直接参与梯度计算；其次对梯度图像进行细化和二值化以提取边缘图像。实验证明，本文算法对脉冲噪声污染图像边缘检测效果良好，较好地抑制了脉冲噪声的影响，而且提取的图像边缘较细，轮廓清晰。和传统的边缘检测算法及基于小波模变换的边缘检测算法相比，算法在抑噪能力上和边缘提取效果上均比较优秀。
제출료일충침대맥충조성도상적변연검측산법，산법차감료중치려파적사상，병채용십자형권적모판계산도상제도。수선，대삼여도상중제도계산적상소점진행역치판단，여과시조성점，해점상소치용3x3창구중치려파결과치체대，연후삼여제도계산，여과불시조성점칙직접삼여제도계산；기차대제도도상진행세화화이치화이제취변연도상。실험증명，본문산법대맥충조성오염도상변연검측효과량호，교호지억제료맥충조성적영향，이차제취적도상변연교세，륜곽청석。화전통적변연검측산법급기우소파모변환적변연검측산법상비，산법재억조능력상화변연제취효과상균비교우수。
The paper proposed an edge detection algorithm aiming at impulse noise image, the algorithm references traditional median filter idea, and uses cross convolution template to calculate image gradient. First, takes judgment for each point attending gradient calculation in image by threshold, if it is noise point, its pixel value will be replaced by result of median filter with 3x3 windows, and then participates in the gradient calculation. If the point is non-noise point, will directly participates calculation with its pixel value. Second,we refine the gradient image, and extract the image edge. Experiments show that the new algorithm has good detection effect and suppressing noise performance, the image edge of exaction is fine, the outline is clear. Compared with the traditional edge detection algorithm and wavelet transform edge detection algorithm, this algorithm is relatively good on suppression noise ability and edge effect.