【Python】用Python将图片转换成Vivado Rom初始化coe文件

1. 介绍

使用XIlinx开发平台Vivado玩FPGA时，需要将一种图片转换成.coe文件用于初始化Rom IP。本文用Python编写一个脚本实现图片转换成Vivado Rom初始化coe文件。如下图分别生成RGB三个coe文件。

2. 程序介绍

2.1 读取图片并调整图片大小

有时图片大小不合适需要调整到指定像素比。通过设置set_size、img_w、img_h进行改变。

import numpy as np 
from PIL import Image 

# 图片文件路径
img_path = 'Dataset/Lena.png'

# 调整原图像大小可设置set_size=1,反之为=0
set_size = 0
img_w = 300 
img_h = 200 

# 读取图片并做大小调整
def gen_img(): 
    img = Image.open(img_path)  
    if set_size == 1: 
        img = img.resize((img_w, img_h), Image.ANTIALIAS)
        img.save('re_picture.jpg')
    img = np.array(img)

    return img

2.2 生成coe文件

# 分别生成R G B的coe文件
def gen_coe(img):
    img_R = []
    img_G = []
    img_B = []
    for i in range(img.shape[0]): 
        for j in range(img.shape[1]): 
            img_R.append(img[i][j][0])
            img_G.append(img[i][j][1])
            img_B.append(img[i][j][2])

    with open('Picture_R_Rom.coe', 'w') as f:
        f.writelines('memory_initialization_radix = 10;\nmemory_initialization_vector = ')
        for i in range(len(img_R)): 
            if i % img.shape[1] == 0: 
                f.write('\n')
            f.write(str(img_R[i]).rjust(4) + ',')
    f.close()

    with open('Picture_G_Rom.coe', 'w') as f:
        f.writelines('memory_initialization_radix = 10;\nmemory_initialization_vector = ')
        for i in range(len(img_G)): 
            if i % img.shape[1] == 0: 
                f.write('\n')
            f.write(str(img_G[i]).rjust(4) + ',')
    f.close()

    with open('Picture_B_Rom.coe', 'w') as f:
        f.writelines('memory_initialization_radix = 10;\nmemory_initialization_vector = ')
        for i in range(len(img_B)): 
            if i % img.shape[1] == 0: 
                f.write('\n')
            f.write(str(img_B[i]).rjust(4) + ',')
    f.close()

2.3 coe文件数据转换成图片

将coe文件反向转换成图片

# coe文件转换成图片
def coe2picture(): 
    infos_R = []
    with open('Picture_R_Rom.coe', 'r') as f:
        info = f.read().splitlines()
        for i in range(2, len(info)): 
            info1 = [int(a) for a in info[i].split(',')[:-1]]
            infos_R.append(info1)
    f.close()

    infos_G = []
    with open('Picture_G_Rom.coe', 'r') as f:
        info = f.read().splitlines()
        for i in range(2, len(info)): 
            info1 = [int(a) for a in info[i].split(',')[:-1]]
            infos_G.append(info1)
    f.close()

    infos_B = []
    with open('Picture_B_Rom.coe', 'r') as f:
        info = f.read().splitlines()
        for i in range(2, len(info)): 
            info1 = [int(a) for a in info[i].split(',')[:-1]]
            infos_B.append(info1)
    f.close()

    img_array = np.empty((np.array(infos_B).shape[0],np.array(infos_B).shape[1],3), dtype=int)
    if set_size == 1: 
        img_array = np.empty((img_h,img_w,3), dtype=int)

    for i in range(np.array(infos_B).shape[0]): 
        for j in range(np.array(infos_B).shape[1]):
            img_array[i][j][0] = infos_R[i][j]
            img_array[i][j][1] = infos_G[i][j]
            img_array[i][j][2] = infos_B[i][j]


    img_pil = Image.fromarray(np.uint8(img_array))
    img_pil.save('coe2picture.jpg')

3. 案例实践

完整程序：

import numpy as np 
from PIL import Image 

# 图片文件路径
img_path = 'Dataset/Lena.png'

# 调整原图像大小可设置set_size=1,反之为=0
set_size = 0
img_w = 300 
img_h = 200 

# 读取图片并做大小调整
def gen_img(): 
    img = Image.open(img_path)  
    if set_size == 1: 
        img = img.resize((img_w, img_h), Image.ANTIALIAS)
        img.save('re_picture.jpg')
    img = np.array(img)

    return img

# 分别生成R G B的coe文件
def gen_coe(img):
    img_R = []
    img_G = []
    img_B = []
    for i in range(img.shape[0]): 
        for j in range(img.shape[1]): 
            img_R.append(img[i][j][0])
            img_G.append(img[i][j][1])
            img_B.append(img[i][j][2])

    with open('Picture_R_Rom.coe', 'w') as f:
        f.writelines('memory_initialization_radix = 10;\nmemory_initialization_vector = ')
        for i in range(len(img_R)): 
            if i % img.shape[1] == 0: 
                f.write('\n')
            f.write(str(img_R[i]).rjust(4) + ',')
    f.close()

    with open('Picture_G_Rom.coe', 'w') as f:
        f.writelines('memory_initialization_radix = 10;\nmemory_initialization_vector = ')
        for i in range(len(img_G)): 
            if i % img.shape[1] == 0: 
                f.write('\n')
            f.write(str(img_G[i]).rjust(4) + ',')
    f.close()

    with open('Picture_B_Rom.coe', 'w') as f:
        f.writelines('memory_initialization_radix = 10;\nmemory_initialization_vector = ')
        for i in range(len(img_B)): 
            if i % img.shape[1] == 0: 
                f.write('\n')
            f.write(str(img_B[i]).rjust(4) + ',')
    f.close()

# coe文件转换成图片
def coe2picture(): 
    infos_R = []
    with open('Picture_R_Rom.coe', 'r') as f:
        info = f.read().splitlines()
        for i in range(2, len(info)): 
            info1 = [int(a) for a in info[i].split(',')[:-1]]
            infos_R.append(info1)
    f.close()

    infos_G = []
    with open('Picture_G_Rom.coe', 'r') as f:
        info = f.read().splitlines()
        for i in range(2, len(info)): 
            info1 = [int(a) for a in info[i].split(',')[:-1]]
            infos_G.append(info1)
    f.close()

    infos_B = []
    with open('Picture_B_Rom.coe', 'r') as f:
        info = f.read().splitlines()
        for i in range(2, len(info)): 
            info1 = [int(a) for a in info[i].split(',')[:-1]]
            infos_B.append(info1)
    f.close()

    img_array = np.empty((np.array(infos_B).shape[0],np.array(infos_B).shape[1],3), dtype=int)
    if set_size == 1: 
        img_array = np.empty((img_h,img_w,3), dtype=int)

    for i in range(np.array(infos_B).shape[0]): 
        for j in range(np.array(infos_B).shape[1]):
            img_array[i][j][0] = infos_R[i][j]
            img_array[i][j][1] = infos_G[i][j]
            img_array[i][j][2] = infos_B[i][j]


    img_pil = Image.fromarray(np.uint8(img_array))
    img_pil.save('coe2picture.jpg')

if __name__ == '__main__': 
    img = gen_img()
    gen_coe(img)
    coe2picture()

运行之后可分别得到RGB三个coe文件和一张coe转图片反向生成的新图片。如果设置了调整图片大小还会生成一个调整图片大小后的图片。