区域生长是一种串行区域分割的图像分割方法。区域生长是指从某个像素出发,按照一定的准则,逐步加入邻近像素,当满足一定的条件时,区域生长终止。区域生长的好坏决定于1.初始点(种子点)的选取。2.生长准则。3.终止条件。区域生长是从某个或者某些像素点出发,最后得到整个区域,进而实现目标的提取。
区域生长的原理:
区域生长的基本思想是将具有相似性质的像素集合起来构成区域。具体先对每个需要分割的区域找一个种子像素作为生长起点,然后将种子像素和周围邻域中与种子像素有相同或相似性质的像素(根据某种事先确定的生长或相似准则来判定)合并到种子像素所在的区域中。将这些新像素当作新的种子继续上面的过程,直到没有满足条件的像素可被包括进来。这样一个区域就生长成了。
区域生长实现的步骤如下:
1. 对图像顺序扫描!找到第1个还没有归属的像素, 设该像素为(x0, y0);
2. 以(x0, y0)为中心, 考虑(x0, y0)的4邻域像素(x, y)如果(x0, y0)满足生长准则, 将(x, y)与(x0, y0)合并(在同一区域内), 同时将(x, y)压入堆栈;
3. 从堆栈中取出一个像素, 把它当作(x0, y0)返回到步骤2;
4. 当堆栈为空时!返回到步骤1;
5. 重复步骤1 - 4直到图像中的每个点都有归属时。生长结束。
二维平面图像
import numpy as np
import cv2
class Point(object):
def __init__(self,x,y):
self.x = x
self.y = y
def getX(self):
return self.x
def getY(self):
return self.y
def getGrayDiff(img,currentPoint,tmpPoint):
return abs(int(img[currentPoint.x,currentPoint.y]) - int(img[tmpPoint.x,tmpPoint.y]))
def selectConnects(p):
if p != 0:
connects = [Point(-1, -1), Point(0, -1), Point(1, -1), Point(1, 0), Point(1, 1), \
Point(0, 1), Point(-1, 1), Point(-1, 0)]
else:
connects = [ Point(0, -1), Point(1, 0),Point(0, 1), Point(-1, 0)]
return connects
def regionGrow(img,seeds,thresh,p = 1):
height, weight = img.shape
seedMark = np.zeros(img.shape)
seedList = []
for seed in seeds:
seedList.append(seed)
label = 1
connects = selectConnects(p)
while(len(seedList)>0):
currentPoint = seedList.pop(0)
seedMark[currentPoint.x,currentPoint.y] = label
for i in range(8):
tmpX = currentPoint.x + connects[i].x
tmpY = currentPoint.y + connects[i].y
if tmpX < 0 or tmpY < 0 or tmpX >= height or tmpY >= weight:
continue
grayDiff = getGrayDiff(img,currentPoint,Point(tmpX,tmpY))
if grayDiff < thresh and seedMark[tmpX,tmpY] == 0:
seedMark[tmpX,tmpY] = label
seedList.append(Point(tmpX,tmpY))
return seedMark
img = cv2.imread('lean.png',0)
seeds = [Point(10,10),Point(82,150),Point(20,300)]
binaryImg = regionGrow(img,seeds,10)
cv2.imshow(' ',binaryImg)
cv2.waitKey(0)
三维体素数据:
import numpy as np
def grow(img, seed, t):
"""
img: ndarray, ndim=3
An image volume.
seed: tuple, len=3
Region growing starts from this point.
t: int
The image neighborhood radius for the inclusion criteria.
"""
seg = np.zeros(img.shape, dtype=np.bool)
checked = np.zeros_like(seg)
seg[seed] = True
checked[seed] = True
needs_check = get_nbhd(seed, checked, img.shape)
while len(needs_check) > 0:
pt = needs_check.pop()
# Its possible that the point was already checked and was
# put in the needs_check stack multiple times.
if checked[pt]: continue
checked[pt] = True
# Handle borders.
imin = max(pt[0]-t, 0)
imax = min(pt[0]+t, img.shape[0]-1)
jmin = max(pt[1]-t, 0)
jmax = min(pt[1]+t, img.shape[1]-1)
kmin = max(pt[2]-t, 0)
kmax = min(pt[2]+t, img.shape[2]-1)
if img[pt] >= img[imin:imax+1, jmin:jmax+1, kmin:kmax+1].mean():
# Include the voxel in the segmentation and
# add its neighbors to be checked.
seg[pt] = True
needs_check += get_nbhd(pt, checked, img.shape)
return seg
区域生长涉及种子选取,提供一个获取图像zuo'坐标的函数:
def on_mouse(event, x,y, flags , params):
if event == cv2.EVENT_LBUTTONDOWN:
print('Seed: ' + 'Point' + '('+str(x) + ', ' + str(y)+')', imger[y, x])
clicks.append((y, x))
cv2.setMouseCallback('input', on_mouse, 0, )
‘input’是你显示图像的命名。