用python做图像处理。有些特殊需求需要用双重for循环遍历图像来操作例如下面代码
def getbinarizationimg(simg, targeth, targetw):
print(simg.shape)
h,w,c = simg.shape
box = np.zeros((h, w),dtype=np.uint8)
pole = np.zeros((h, w),dtype=np.uint8)
for u in range(h):
for v in range(w):
if simg[u][v][0] == 0 and simg[u][v][1] == 220 and simg[u][v][2] == 220:
box[u][v] = 255
img_box = cv2.resize(box, (targetw, targeth), fx=0, fy=0,
interpolation=cv2.INTER_NEAREST)
return img_box
图像大小为1024 * 2048 .主函数中测试结果如下
elapse_time = 3.5969557762145996
elapse_time = 3.608659029006958
elapse_time = 3.667614459991455
elapse_time = 3.546481132507324
直接疯掉的节奏啊。后来看到KCF算法中有个加速方法,拿来用用
from numba import jit
@jit()
def getbinarizationimg(simg, targeth, targetw):
print(simg.shape)
h,w,c = simg.shape
box = np.zeros((h, w),dtype=np.uint8)
pole = np.zeros((h, w),dtype=np.uint8)
for u in range(h):
for v in range(w):
if simg[u][v][0] == 0 and simg[u][v][1] == 220 and simg[u][v][2] == 220:
box[u][v] = 255
img_box = cv2.resize(box, (targetw, targeth), fx=0, fy=0,
interpolation=cv2.INTER_NEAREST)
return img_box
elapse_time = 0.29776477813720703
elapse_time = 0.0017366409301757812
elapse_time = 0.001975536346435547
elapse_time = 0.001725912094116211
elapse_time = 0.0018794536590576172
性能差不少啊。