以遮罩(mask处理影像.ppt

Sharpening Spatial Filter Student: Pei - Jun Jiang Advisor: Ku-Yaw Chang Outline Introduction Derivative operators Soble Laplacian Introduction Spatial filtering 以遮罩(mask)處理影像 或稱濾波器(filter)、濾波器遮罩(filter mask)、核心(kernel)、模板(template) 遮罩運算 輸出像素值 = 利用遮罩運算於相對位置鄰近區之所有像素 鄰近區大小即為遮罩大小，並以由左往右、由上往下方式滑動 每個輸出像素均以相同方式運算 超出影像範圍之鄰近區：補 0 或重複邊緣像素值 Introduction Smoothing Spatial Filters 將影像模糊化或減少雜訊 用於影像的前處理 去除影像中不必要的細部，並填補影像中線段或曲線間之斷層 Sharpening Spatial Filters 影像銳利化及增強對比 突顯影像中細微的部分 增強因錯誤或利用特殊方法取得影像而造成模糊的細部 強調影像中物體的輪廓 Outline Introduction Derivative operators Laplacian Gradient Derivative operators Some sharpening filters are based on First order derivative Second order derivative Interested in the behavior of Flat segment Ramp Step Used to model Noise points, lines or edge Derivative operators First order derivative Second order derivative Derivative operators Derivative operators First derivative Must be zero in flat segment Must be nonzero at the onset of a gray-level step or ramps Must be nonzero along ramps Second derivative Must be zero in flat area Must be nonzero at the onset and end of a gray-level step or ramps Must be zero along ramps Derivative operators Conclusion First derivative Produce thicker edge in an image Have stronger response to a gray lever step Second derivative Have stronger response to fine detail Produce a double response at step changes in gray level Outline Introduction Derivative operators Sobel Laplacian Sobel First order derivative 有方向性 邊緣偵測 常用於機械視覺 判別方法 經過運算元計算後，若在門檻值內且梯度相近、方向向量相近，即判定為邊緣 Sobel 設A為原始圖像 橫向 Gx = * A = Ex 縱向 Gy = * A = Ey T : 門檻值 E : 強度 Θ: 方向 當E > T 時，表該像素為邊緣點 Sobel Outline Introduction Derivative operators Sobel Laplacian Laplacian Second order derivative 對雜訊相當敏感 對邊緣方向具不變性 無法檢測出邊緣的方向 邊緣強化 將影像中，峰值部份或是灰階值最高的部份，再提高其比重。 Laplacian Linear operator X - direction Summing Y - direction Laplacian Masks Laplacian Thank