DOG : difference of Gaussians

-
For 2D case：
 

大部分的边缘锐化算子使用增强高频信号的方法，但是因为随机噪声也是高频信号，很多锐化算子也增强了噪声。DOG算法去除的高频信号中通常包含了随机噪声，所以这种方法是最适合处理那些有高频噪声的图像。这个算法的一个主要缺点就是在调整图像对比度的过程中信息量会减少。
当它被用于图像增强时，DOG算法中两个高斯核的半径之比通常为4:1或5:1。当设为1.6时，即为高斯拉普拉斯算子的近似。高斯拉普拉斯算子在多尺度多分辨率像片。用于近似高斯拉普拉斯算子两个高斯核的确切大小决定了两个高斯模糊后的影像间的尺度。
DOG也被用于尺度不变特征变换中的斑点检测。事实上，DOG算法作为两个多元正态分布的差通常总额为零，把它和一个恒定信号进行卷积没有意义。当K约等于1.6时它很好的近似了高斯拉普拉斯变换，当K约等于5时又很好的近似了视网膜上神经节细胞的视野。它可以很好的作为一个实时斑点检测算子和尺度选择算子应用于递归程序。


留言

張貼留言

這個網誌中的熱門文章

AndrewNg's CNN notes(practical issues)

-
圖片
In this example, the author adapted the architecture obtained from training ILSVRC-2011 dataset which consists of 1000 classes with approximately 1.5 million images. And using the pre-trained VGG-16 network, the author fine-tuned the network on PASCAL VOC detection data (20 object classes, and 1 background class) and replaced the original 1000 way classification layer by a randomly initialized 21-way classification layer. Apart from that, the author keep the rest of the CNN architecture remains unchanged. After training of CNN parameters and filter the region proposals based on their IoU value, at the test stage, RCNN uses Selective Search to extract ~300 boxes that likely contain objects and evaluates the ConvNet on each one of them, followed by NMS(non-maximum suppression) within each class.
閱讀完整內容

New findings in Numpy

-
1, Ellipsis >>> a = np.arange(6).reshape(2,3) >>> a array([[0, 1, 2], [3, 4, 5]]) >>> for x in np.nditer(a, op_flags=['readwrite']):                                                       x[...] = 2 * x >>> a array([[ 0, 2, 4], [ 6, 8, 10]]) 2, array or matrix Operator  * ,  dot() , and  multiply() : For  array ,  ‘``*``’ means element-wise multiplication , and the  dot()  function is used for matrix multiplication. For  matrix ,  ‘``*``’ means matrix multiplication , and the  multiply()  function is used for element-wise multiplication. Handling of vectors (one-dimensional arrays) For  array , the  vector shapes 1xN, Nx1, and N are all different things . Operations like  A[:,1]  return a one-dimensional array of shape N, not...
閱讀完整內容