video2video.py

"""
@author: Viet Nguyen <nhviet1009@gmail.com>
"""

import argparse

import cv2
import numpy as np
from PIL import Image, ImageFont, ImageDraw, ImageOps


def get_args():
    parser = argparse.ArgumentParser("Image to ASCII")
    parser.add_argument(
        "--input", type=str, default="data/input.mp4", help="Path to input video"
    )
    parser.add_argument(
        "--output", type=str, default="data/output.mp4", help="Path to output video"
    )
    parser.add_argument(
        "--mode",
        type=str,
        default="simple",
        choices=["simple", "complex"],
        help="10 or 70 different characters",
    )
    parser.add_argument(
        "--background",
        type=str,
        default="white",
        choices=["black", "white"],
        help="background's color",
    )
    parser.add_argument(
        "--num_cols",
        type=int,
        default=100,
        help="number of character for output's width",
    )
    parser.add_argument("--scale", type=int, default=1, help="upsize output")
    parser.add_argument("--fps", type=int, default=0, help="frame per second")
    parser.add_argument(
        "--overlay_ratio", type=float, default=0.2, help="Overlay width ratio"
    )
    args = parser.parse_args()
    return args


def main(opt):
    if opt.mode == "simple":
        CHAR_LIST = r"@%#*+=-:. "
    else:
        CHAR_LIST = (
            r"$@B%8&WM#*oahkbdpqwmZO0QLCJUYXzcvunxrjft/\|()1{}[]?-_+~<>i!lI;:,\"^`'. "
        )
    if opt.background == "white":
        bg_code = 255
    else:
        bg_code = 0
    font = ImageFont.truetype("fonts/DejaVuSansMono-Bold.ttf", size=int(10 * opt.scale))
    cap = cv2.VideoCapture(opt.input)
    if opt.fps == 0:
        fps = int(cap.get(cv2.CAP_PROP_FPS))
    else:
        fps = opt.fps
    num_chars = len(CHAR_LIST)
    num_cols = opt.num_cols
    while cap.isOpened():
        flag, frame = cap.read()
        if flag:
            image = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
        else:
            break
        height, width = image.shape
        cell_width = width / opt.num_cols
        cell_height = 2 * cell_width
        num_rows = int(height / cell_height)
        if num_cols > width or num_rows > height:
            print("Too many columns or rows. Use default setting")
            cell_width = 6
            cell_height = 12
            num_cols = int(width / cell_width)
            num_rows = int(height / cell_height)
        bbox = font.getbbox("A")
        char_width = bbox[2] - bbox[0]
        char_height = bbox[3] - bbox[1]
        out_width = char_width * num_cols
        out_height = 2 * char_height * num_rows
        out_image = Image.new("L", (out_width, out_height), bg_code)
        draw = ImageDraw.Draw(out_image)
        for i in range(num_rows):
            line = (
                "".join(
                    [
                        CHAR_LIST[
                            min(
                                int(
                                    np.mean(
                                        image[
                                            int(i * cell_height) : min(
                                                int((i + 1) * cell_height), height
                                            ),
                                            int(j * cell_width) : min(
                                                int((j + 1) * cell_width), width
                                            ),
                                        ]
                                    )
                                    * num_chars
                                    / 255
                                ),
                                num_chars - 1,
                            )
                        ]
                        for j in range(num_cols)
                    ]
                )
                + "\n"
            )
            draw.text((0, i * char_height), line, fill=255 - bg_code, font=font)

        if opt.background == "white":
            cropped_image = ImageOps.invert(out_image).getbbox()
        else:
            cropped_image = out_image.getbbox()
        out_image = out_image.crop(cropped_image)
        out_image = cv2.cvtColor(np.array(out_image), cv2.COLOR_GRAY2BGR)
        out_image = np.array(out_image)
        try:
            out
        except:
            out = cv2.VideoWriter(
                opt.output,
                cv2.VideoWriter_fourcc(*"XVID"),
                fps,
                ((out_image.shape[1], out_image.shape[0])),
            )

        if opt.overlay_ratio:
            height, width, _ = out_image.shape
            overlay = cv2.resize(
                frame, (int(width * opt.overlay_ratio), int(height * opt.overlay_ratio))
            )
            out_image[
                height - int(height * opt.overlay_ratio) :,
                width - int(width * opt.overlay_ratio) :,
                :,
            ] = overlay
        out.write(out_image)
    cap.release()
    out.release()


if __name__ == "__main__":
    opt = get_args()
    main(opt)