這兩天，我在製作教學短視頻時，想將一些關鍵幀（Key Frame）快速製作成影片格式，好讓我快速就著關鍵資訊進行錄音，然後就可以快速後製。

我搜尋了在 Python 如何將圖片轉變成影片，找到 OpenCV 庫可以達成，找到幾個網站都差不多，代碼大致如下：

import cv2
import glob

frameSize = (2266, 1488)

fourcc = cv2.VideoWriter_fourcc(*'mp4v')

out = cv2.VideoWriter('output_video.mp4',fourcc, 0.25, frameSize)

for filename in glob.glob('./test-img-to-videos/*.jpg'):
    img = cv2.imread(filename)
    print(img.shape)
    out.write(img)

out.release()

來源：

當中有兩點注意：

生成影片用的每幀圖片尺寸必需一致。
我找到幾個網上代碼都是生成 AVI。我改為使用 *'mp4v' 格式以生成 MP4 格式。

再經過調整，我作了以下更改：

取得第一張圖片的尺寸作為影片尺寸。
判斷其餘圖片自動符合影片尺寸，自動縮放置中。（使用了 Pillow 庫）
對 glob 的檔案列表進行排序，按名稱順序排。（https://stackoverflow.com/a/6774404）

新改善後代碼：

import cv2
import glob
from PIL import Image
import numpy as np

def image_to_frame(filename, frame_size=(1920,1080), fit=True, background=(0,0,0)):
    im = Image.new('RGB', frame_size, background)
    img = Image.open(filename)
    if not fit:
        im.paste(img)
    else:
        width, height = img.size
        frame_width, frame_height = frame_size
        ratio = min(frame_width/width, frame_height/height)
        img = img.resize((round(width*ratio), round(height*ratio)))
        pos_x = round((frame_width - width*ratio)/2)
        pos_y = round((frame_height - height*ratio)/2)
        im.paste(img, (pos_x, pos_y))
    return im

def cv2_image_from_pillow(pillow_image):
    """Converting from Pillow image to cv2 image"""
    # https://medium.com/analytics-vidhya/the-ultimate-handbook-for-opencv-pillow-72b7eff77cd7
    cv2_img = np.array(pillow_image)
    return cv2.cvtColor(cv2_img, cv2.COLOR_RGB2BGR)

frame_rate = 0.25
frame_size = None
# frame_size = (1920,1080) # if you want a pre-defined frame size.

frames = []

files = glob.glob('./test-img-to-videos/*.jpg')
files.sort()

for filename in files:
    img = cv2.imread(filename)
    height, width, _ = img.shape
    if frame_size == None:
        frame_size = [width, height]
    elif (width, height) != frame_size:
        img = image_to_frame(filename, frame_size=frame_size)
        img = cv2_image_from_pillow(img)
    frames.append(img)

fourcc = cv2.VideoWriter_fourcc(*'mp4v')
out = cv2.VideoWriter('output_video.mp4',fourcc, frame_rate, frame_size)

for frame in frames:
    out.write(frame)

out.release()

當中，輸入資料夾及輸出檔案名仍為 hardcoded 固定名稱，若進一步要變成指令（Command Line Interface）來執行，可以將輸入輸出名稱改以 args 來存取。

例如可以通過 python img_to_mp4.py source_images output_keyframes.mp4 來生成。

最終代碼：

import cv2
import glob
from PIL import Image
import numpy as np

def image_to_frame(filename, frame_size=(1920,1080), fit=True, background=(0,0,0)):
    im = Image.new('RGB', frame_size, background)
    img = Image.open(filename)
    if not fit:
        im.paste(img)
    else:
        width, height = img.size
        frame_width, frame_height = frame_size
        ratio = min(frame_width/width, frame_height/height)
        img = img.resize((round(width*ratio), round(height*ratio)))
        pos_x = round((frame_width - width*ratio)/2)
        pos_y = round((frame_height - height*ratio)/2)
        im.paste(img, (pos_x, pos_y))
    return im

def cv2_image_from_pillow(pillow_image):
    """Converting from Pillow image to cv2 image"""
    # https://medium.com/analytics-vidhya/the-ultimate-handbook-for-opencv-pillow-72b7eff77cd7
    cv2_img = np.array(pillow_image)
    return cv2.cvtColor(cv2_img, cv2.COLOR_RGB2BGR)


def main():
    import sys

    if len(sys.argv) != 3:
        print("Please provide two arguments: source directory (*.jpg) and output_video (.mp4) name.")
        return

    source_dir = sys.argv[1]
    output_video = sys.argv[2]

    frame_rate = 0.25
    frame_size = None
    # frame_size = (1920,1080) # if you want a pre-defined frame size.

    frames = []

    files = glob.glob(f"./{source_dir}/*.jpg")
    files.sort()

    for filename in files:
        img = cv2.imread(filename)
        height, width, _ = img.shape
        if frame_size == None:
            frame_size = [width, height]
        elif (width, height) != frame_size:
            img = image_to_frame(filename, frame_size=frame_size)
            img = cv2_image_from_pillow(img)
        frames.append(img)

    fourcc = cv2.VideoWriter_fourcc(*'mp4v')
    out = cv2.VideoWriter(output_video,fourcc, frame_rate, frame_size)

    for frame in frames:
        out.write(frame)

    out.release()

    print(f"Done creating video {output_video} from folder {source_dir}.")

if __name__ == "__main__":
    main()