METHOD OF ESTIMATION OF HISTOGRAM-OF-ORIENTED GRADIENTS

摘要

An image processing method adapted to detect and identify various classes of objects in images, which uses a Histogram of Gradients (HOG) methodology to determine gradient orientation in portions of an image; wherein in the HOG methodology, the length (M) of the gradient vectors associated with particular pixels are determined from pre-calculated components of the vector in x and y coordinates/directions (x,y), using the following formula:$M=\max \left(\left|x\right|,\left|y\right|\right)+\sum _{i=0}^k{s}_i\cdot b_i\cdot \left[\min \left(\left|x\right|,\left|y\right|\right)/2^{i+2}\right]\mathop{\mathrm{image}}$ where max(|x|, |y|) is the absolute value of either said x or y components, whichever is the larger;[min(|x|, |y|)/2 is the absolute value of either x or y components, whichever is the smaller; k is an integer value; and s_i and b_i are integers having values of 0, 1 or -1