我帮你排查了代码里的核心问题，主要是坐标处理的精度误差导致切片拼接错位，下面是具体分析和解决方法：

问题根源分析 #

• 浮点数坐标取整错误：你把x和y直接用round()取整，但原切片的定位坐标是精确的浮点数（比如181.5px），哪怕0.5px的偏差，多个切片叠加后就会出现重叠、内容缺失的问题。
• 画布尺寸计算精度不足：当前对x+w取round后再求最大值，会导致画布边缘尺寸不够，可能截断部分内容。
• 未验证图像实际尺寸：文件名标注的宽高可能和下载到的图像实际尺寸有出入，直接按标注坐标粘贴也会引发错位。

修正后的代码 #

针对这些问题，调整后的代码如下：

import requests
from PIL import Image
from io import BytesIO
import math

urls = [
    "https://epaper.janayugomonline.com/assets/epaper/thiruvananthapuram/2026/04/05/slices/01_38_43_141.5_359.jpg",
    "https://epaper.janayugomonline.com/assets/epaper/thiruvananthapuram/2026/04/05/slices/01_181.5_41_738_148.jpg",
    "https://epaper.janayugomonline.com/assets/epaper/thiruvananthapuram/2026/04/05/slices/01_38.5_406_137_180.jpg",
    "https://epaper.janayugomonline.com/assets/epaper/thiruvananthapuram/2026/04/05/slices/01_36.5_579_147_359.jpg",
    "https://epaper.janayugomonline.com/assets/epaper/thiruvananthapuram/2026/04/05/slices/01_182.5_184_739_283.jpg",
    "https://epaper.janayugomonline.com/assets/epaper/thiruvananthapuram/2026/04/05/slices/01_182.5_466_592_474.jpg",
    "https://epaper.janayugomonline.com/assets/epaper/thiruvananthapuram/2026/04/05/slices/01_774.5_465_143_263.jpg",
    "https://epaper.janayugomonline.com/assets/epaper/thiruvananthapuram/2026/04/05/slices/01_773.5_725_144_215.jpg",
    "https://epaper.janayugomonline.com/assets/epaper/thiruvananthapuram/2026/04/05/slices/01_39.5_938_440_310.jpg",
    "https://epaper.janayugomonline.com/assets/epaper/thiruvananthapuram/2026/04/05/slices/01_38.5_1242_440_147.jpg",
    "https://epaper.janayugomonline.com/assets/epaper/thiruvananthapuram/2026/04/05/slices/01_477.5_937_151_458.jpg",
    "https://epaper.janayugomonline.com/assets/epaper/thiruvananthapuram/2026/04/05/slices/01_625.5_937_294_451.jpg"
]

tiles = []
max_w, max_h = 0.0, 0.0

for url in urls:
    name = url.split("/")[-1].replace(".jpg", "")
    _, x, y, w, h = name.split("_")
    x, y, w, h = map(float, [x, y, w, h])
    
    # 下载并读取图像
    img_response = requests.get(url)
    img = Image.open(BytesIO(img_response.content))
    
    # 验证图像实际尺寸与标注是否一致（可选但推荐）
    actual_w, actual_h = img.size
    if abs(actual_w - w) > 1 or abs(actual_h - h) > 1:
        print(f"Warning: Image {name} has actual size ({actual_w}, {actual_h}) vs labeled ({w}, {h})")
    
    # 用int()直接截断小数，贴合CSS亚像素渲染的实际定位逻辑
    tiles.append((img, int(x), int(y)))
    
    # 精确计算画布的最大边界
    max_w = max(max_w, x + w)
    max_h = max(max_h, y + h)

# 用ceil确保画布能完整容纳所有切片
canvas_width = math.ceil(max_w)
canvas_height = math.ceil(max_h)
canvas = Image.new("RGB", (canvas_width, canvas_height), (255, 255, 255))

for img, x, y in tiles:
    canvas.paste(img, (x, y))

canvas.save("output.jpg")

关键调整说明 #

1. 坐标处理：用int()直接截断浮点数坐标的小数部分，而不是round()，更贴合原页面CSS的亚像素渲染逻辑，避免累积偏移。
2. 画布尺寸：保留浮点数计算的最大宽高，最后用math.ceil()确保画布能容纳所有切片的边缘内容，不会截断。
3. 尺寸验证：新增了图像实际尺寸与标注尺寸的校验，如果出现不符可以及时排查下载或标注错误。

内容的提问来源于stack exchange，提问作者nicholaspooran