一般情况下,有python的三方库的软件,可以读RGB二进制的Raw数据,生成Numpy数据。比如说OpenCV之类的。但是,我们也可以使用python的裸代码,得到Numpy。
知识点:
- numpy.frombuffer 将bytes格式转换为numpy ndarray
- numpy.reshape 可以将一维的numpy数组转换成三维的数组
- numpy.dstack 可以将三维的RBA数据扩充成RGBA 数据
- numpy.concatenate 可以扩充到更高维度的numpy 数据
代码例子如下:
#!/usr/bin/env python
# encoding: utf-8
import os
import numpy as np
def read_raw_to_rgb(raw_path, width, height, channel):
""" 读取raw数据转换成 RGB 的numpy 格式的数据 """
assert os.path.exists(raw_path), r'输入的文件必须存在'
raw_data = open(raw_path, 'rb').read()
raw_data_shape = (width, height, channel)
np_raw_data = np.frombuffer(raw_data, dtype=np.uint8)
reshaped_data = np_raw_data.reshape(raw_data_shape)
return reshaped_data
def convert_rgb_to_rgba(rgb_data, alpha):
""" 将 RGB 数据扩充成 RGBA 的数据"""
r = rgb_data[:, :, 0]
g = rgb_data[:, :, 1]
b = rgb_data[:, :, 2]
a = np.zeros(r.shape, np.uint8)
a[:, :] = alpha
rgba_data = np.dstack((r, g, b, a))
return rgba_data
def create_alpha_data_according_rgb_data(rgb_data, alpha):
""" 根据rgb_data的维度创建alpha_data"""
r = rgb_data[:, :, 0]
a = np.zeros(r.shape, np.double)
a[:, :] = alpha
return a
vr_path = r'D:\temp\result\vr_512_512.raw'
sr_path = r'D:\temp\result\sr_512_512.raw'
vr_rgb_data = read_raw_to_rgb(vr_path, 512, 512, 3)
vr_alpha_data = create_alpha_data_according_rgb_data(vr_rgb_data, 0.1)
sr_rgb_data = read_raw_to_rgb(sr_path, 512, 512, 3)
sr_alpha_data = create_alpha_data_according_rgb_data(sr_rgb_data, 0.9)
print(512*512*4)
vr_rgb_data_image_idx = vr_rgb_data[np.newaxis, :, :, :]
sr_rgb_data_image_idx = sr_rgb_data[np.newaxis, :, :, :]
rgb_data_array = np.concatenate((vr_rgb_data_image_idx, sr_rgb_data_image_idx), axis=0)
print(rgb_data_array.shape)
vr_alpha_data_image_idx = vr_alpha_data[np.newaxis, :, :]
sr_alpha_data_image_idx = sr_alpha_data[np.newaxis, :, :]
alpha_data_array = np.concatenate((vr_alpha_data_image_idx, sr_alpha_data_image_idx), axis=0)
print(alpha_data_array.shape)
测试环境中的图像,可以使用 OpenCV读取一张图, 将numpy保存到文件中。