前言

因為最近在搞畢業論文的事情，要做出一下圖像細節對比圖，所以我這里寫了兩個腳本，一個用于框選并同時預覽圖像放大細節，可顯示并返回框選圖像的坐標，另外一個是輸入框選圖像的坐標并將放大的細節放置在圖像中，效果如下所示：

效果也是相當不錯的，好了咱們也不必多說，就是教會大家怎么使用這兩個腳本就可以了。?

框選圖像并預覽放大細節

我們這里寫了一個圖像區域的選擇工具，主要是選擇好圖像路徑，框選和文字的顏色，以及放大的倍數，此處放大的倍數僅用于查看，所以不用擔心最后的效果。

import cv2def select_roi_region(image_path, line_color=(0, 255, 0), zoom_factor=3):drawing = Falseix, iy = -1, -1x, y, w, h = 0, 0, 0, 0img = cv2.imread(image_path)clone = img.copy()# 鼠標回調函數def mouse_callback(event, cur_x, cur_y, flags, param):nonlocal ix, iy, drawing, x, y, w, hif event == cv2.EVENT_LBUTTONDOWN:drawing = Trueix, iy = cur_x, cur_yx, y, w, h = 0, 0, 0, 0elif event == cv2.EVENT_MOUSEMOVE and drawing:temp_img = clone.copy()cv2.rectangle(temp_img, (ix, iy), (cur_x, cur_y), line_color, 2)x1, y1 = min(ix, cur_x), min(iy, cur_y)x2, y2 = max(ix, cur_x), max(iy, cur_y)if x2 > x1 and y2 > y1:try:roi = img[y1:y2, x1:x2]if roi.size > 0:enlarged = cv2.resize(roi, None, fx=3, fy=3,interpolation=cv2.INTER_CUBIC)cv2.imshow("Enlarged Preview", enlarged)except Exception as e:passcur_w = abs(cur_x - ix)cur_h = abs(cur_y - iy)if cur_w > 0 and cur_h > 0:try:roi = img[y1:y2, x1:x2]enlarged = cv2.resize(roi, None, fx=zoom_factor, fy=zoom_factor,interpolation=cv2.INTER_CUBIC)cv2.imshow("Enlarged Preview", enlarged)except:passcv2.putText(temp_img, f"X:{x1} Y:{y1} W:{cur_w} H:{cur_h}",(10, 30), cv2.FONT_HERSHEY_SIMPLEX, 0.7, line_color, 2)cv2.imshow("Select ROI (SPACE=Clear | ENTER=Confirm)", temp_img)elif event == cv2.EVENT_LBUTTONUP:drawing = Falsex = min(ix, cur_x)y = min(iy, cur_y)w = abs(cur_x - ix)h = abs(cur_y - iy)cv2.rectangle(clone, (x, y), (x + w, y + h), line_color, 2)cv2.putText(clone, f"X:{x} Y:{y} W:{w} H:{h}", (10, 30),cv2.FONT_HERSHEY_SIMPLEX, 0.7, line_color, 2)cv2.imshow("Select ROI (SPACE=Clear | ENTER=Confirm)", clone)cv2.namedWindow("Select ROI (SPACE=Clear | ENTER=Confirm)")cv2.setMouseCallback("Select ROI (SPACE=Clear | ENTER=Confirm)", mouse_callback)while True:cv2.imshow("Select ROI (SPACE=Clear | ENTER=Confirm)", clone)key = cv2.waitKey(1) & 0xFF# 空格鍵：清除選擇if key == 32:clone = img.copy()ix, iy = -1, -1x, y, w, h = 0, 0, 0, 0try:cv2.destroyWindow("Enlarged Preview") if cv2.getWindowProperty("Enlarged Preview", 0) >=0 else Noneexcept:passcv2.imshow("Select ROI (SPACE=Clear | ENTER=Confirm)", clone)# 回車鍵：確認選擇if key == 13:try:cv2.destroyWindow("Enlarged Preview")except:passbreakcv2.destroyAllWindows()print(f"Final selection - X:{x} Y:{y} W:{w} H:{h}")return (x, y, w, h)if __name__=="__main__":select_roi_region(r'E:\PythonProject\img_processing_techniques_main\Enlarge_local_details\images\781.png'
)

下面是我們的這個使用效果：

有一點問題就是在繪制好圖像后再選框就會將文字遮擋住：

但是我們是提供了清楚鍵的，你只需要按下空格鍵就可以將全圖的文字和框清理掉了，最后選擇好合適的區域后，按下Enter鍵確定你框選的區域，以便進行下一步操作。

這里返回的坐標是（x，y，w，h），這種方式便于我們控制起始點和框的大小。

框選圖像并放置放大細節

這里需要的是選擇圖像路徑，框選的坐標，也提供放置位置的坐標，放大的系數，線條的顏色，寬度，以及是否繪制箭頭。

如果你不提供放置的位置也可以，我們提供了一種自動計算位置的方法，主要就是比較框選區域的位置，然后計算出其對角線位置返回坐標。

import math
from PIL import Image, ImageDraw, ImageTkdef plot_highlight_region(image_path, region_to_zoom, paste_position=None, zoom_factor=3,line_color="red", line_wide=2, show_arrow=True, arrow_size=5):x, y, w, h = region_to_zoomimg = Image.open(image_path).convert("RGB")img_w, img_h = img.sizeoriginal_copy = img.copy()zoomed_w = int(w * zoom_factor)zoomed_h = int(h * zoom_factor)cropped = original_copy.crop((x, y, x + w, y + h))zoomed = cropped.resize((zoomed_w, zoomed_h), Image.Resampling.LANCZOS)if paste_position is None:if x + w < img_w / 2:paste_x = img_w - zoomed_welse:paste_x = 0if y + h < img_h / 2:paste_y = img_h - zoomed_helse:paste_y = 0paste_x = max(0, min(paste_x, img_w - zoomed_w))paste_y = max(0, min(paste_y, img_h - zoomed_h))paste_position = (paste_x, paste_y)img.paste(zoomed, paste_position)draw = ImageDraw.Draw(img)draw.rectangle([(x, y), (x + w, y + h)],outline=line_color,width=line_wide)paste_x, paste_y = paste_positiondraw.rectangle([paste_position,(paste_x + zoomed_w, paste_y + zoomed_h)],outline=line_color, width=line_wide)if show_arrow:def get_side_center(rect, side):x, y, w, h = rectreturn {'left': (x, y + h // 2),'right': (x + w, y + h // 2),'top': (x + w // 2, y),'bottom': (x + w // 2, y + h)}[side]src_rect = (x, y, w, h)dst_rect = (paste_position[0], paste_position[1], zoomed_w, zoomed_h)dx = (dst_rect[0] + zoomed_w / 2) - (x + w / 2)dy = (dst_rect[1] + zoomed_h / 2) - (y + h / 2)if abs(dx) > abs(dy):src_side = 'right' if dx > 0 else 'left'dst_side = 'left' if dx > 0 else 'right'else:src_side = 'bottom' if dy > 0 else 'top'dst_side = 'top' if dy > 0 else 'bottom'start_point = get_side_center(src_rect, src_side)end_point = get_side_center(dst_rect, dst_side)draw.line([start_point, end_point], fill=line_color, width=line_wide)arrow_size = line_wide * arrow_sizeangle = math.atan2(end_point[1] - start_point[1], end_point[0] - start_point[0])p1 = (end_point[0] - arrow_size * math.cos(angle - math.pi / 6),end_point[1] - arrow_size * math.sin(angle - math.pi / 6))p2 = (end_point[0] - arrow_size * math.cos(angle + math.pi / 6),end_point[1] - arrow_size * math.sin(angle + math.pi / 6))draw.polygon([end_point, p1, p2], fill=line_color)return imgif __name__ == "__main__":# 定義要放大的區域 (x, y, width, height)region_to_zoom = (256, 250, 50, 70)im_path = r'E:\PythonProject\img_processing_techniques_main\Enlarge_local_details\gtimage\781.png'im = plot_highlight_region(im_path, region_to_zoom)im.save("output.png")

我們先來看看，提供了放置坐標的效果：

if __name__ == "__main__":# 定義要放大的區域 (x, y, width, height)region_to_zoom = (256, 250, 50, 70)im_path = r'E:\PythonProject\img_processing_techniques_main\Enlarge_local_details\gtimage\861.png'im = plot_highlight_region(im_path, region_to_zoom, (22, 22))im.save("output.png")

自動計算的效果：

if __name__ == "__main__":# 定義要放大的區域 (x, y, width, height)region_to_zoom = (22, 22, 50, 70)im_path = r'E:\PythonProject\img_processing_techniques_main\Enlarge_local_details\gtimage\861.png'im = plot_highlight_region(im_path, region_to_zoom)im.save("output.png")

當然這里的自動計算還只是四個角。目前來說也算足夠了。

關閉箭頭的效果：

if __name__ == "__main__":# 定義要放大的區域 (x, y, width, height)region_to_zoom = (300, 250, 50, 70)im_path = r'E:\PythonProject\img_processing_techniques_main\Enlarge_local_details\gtimage\861.png'im = plot_highlight_region(im_path, region_to_zoom, show_arrow=False)im.save("output.png")

總結

如果只需要畫框那么直接用下面的簡略版本即可：

def highlight_region(image_path, region_to_zoom, line_color="red", line_wide=2):x, y, w, h = region_to_zoomimg = Image.open(image_path).convert("RGB")img_copy = img.copy()draw = ImageDraw.Draw(img)draw.rectangle([(x, y), (x + w, y + h)],outline=line_color,width=line_wide)return img

我們將前面的兩個腳本組合在一起，便于我們更好的觀察

if __name__ == "__main__":# 定義要放大的區域 (x, y, width, height)from mouse import select_roi_regionim_path = r'E:\PythonProject\img_processing_techniques_main\Enlarge_local_details\gtimage\861.png'region_to_zoom = select_roi_region(im_path)im = plot_highlight_region(im_path, region_to_zoom, show_arrow=False)im.save("output.png")

這里會先運行預選框程序，等按下Enter鍵之后會直接返回坐標。

我們的圖像就生成好了：

寫完這篇我還得繼續去畫圖了。