CAJ | 학술논문

双边滤波亮度相似度因子仅受σr一个参数的约束，很难准确辨析图像平滑区域及细节丰富区域的纹理信息，不能较好地保留纹理细节信息。基于此，提出一种自适应分数阶微分与双边滤波相结合的图像去噪方法。在双边滤波算法的基础上，通过分析局部纹理特征，计算频率变化梯度和幅值变化特征，建立幅值频率非线性指数模型，自适应地选择每个像素点对应的分数阶阶次v,构建分数阶微分掩模算子。实验结果表明，改进的双边滤波算法能够在图像滤波的同时实现纹理细节地保留/增强，获得较好的滤波结果。
쌍변려파량도상사도인자부수σr일개삼수적약속，흔난준학변석도상평활구역급세절봉부구역적문리신식，불능교호지보류문리세절신식。기우차，제출일충자괄응분수계미분여쌍변려파상결합적도상거조방법。재쌍변려파산법적기출상，통과분석국부문리특정，계산빈솔변화제도화폭치변화특정，건립폭치빈솔비선성지수모형，자괄응지선택매개상소점대응적분수계계차v,구건분수계미분엄모산자。실험결과표명，개진적쌍변려파산법능구재도상려파적동시실현문리세절지보류/증강，획득교호적려파결과。
Similar weight of bilateral filtering is only controlled by σr , so the texture detail in smooth area or complex envi-ronment cannot be reserved well. Therefore, a developed bilateral filtering algorithm with adaptive fractional order differ-entiation is proposed. The amplitude-frequency nonlinear model is established through analyzing texture characters and computing frequency gradients/amplitude change based on bilateral filter algorithm, which automatically chooses fractional order v to create mask operator using fractional order differentiation. Experimental results show that the proposed algo-rithm has better performance in noise filtering while maintaining/enhancing texture details robustly.