Python处理图片生成天际线(2024.1.29)

1、天际线简介

天际线SkyLine)顾名思义就是天空与地面的边界线,人站在不同的高度,会看到不同的景色和地平线,天空与地面建筑物分离的标记线,不得不说,每天抬头仰望天空,相信大家都可以看到,它的的确确客观存在,美丽值得欣赏。

2、Python代码

python 复制代码
#-*- coding:utf-8 -*-
import sys
from os.path import exists
import cv2
import numpy as np

def getImage(height, width, channels):
    image = np.zeros([height, width, 3], np.uint8) # 三通道顺序是BGR
    # 三层循环逐个修改像素点
    for row in range(height):
        for col in range(width):
                for c in range(channels):
                    image[row, col, c] = 0
    return image

def isWhite(pixel_value, threshold): #阈值可以取10、20、30、50、100
    res = False
    if pixel_value[0] > threshold and pixel_value[1] > threshold and pixel_value[2] > threshold: # 10、10、10 50、50、50 这里是天空和地面楼山的分界线,需要调参
        res = True
    return res

def isPureWhite(pixel_value):
    res = False
    if pixel_value[0] == 255 and pixel_value[1] == 255 and pixel_value[2] == 255: # >3|>3|>3 10、10、10
        res = True
    return res

def getRowNumberSpecificCol(image, col):
    res_row = -1
    height, width = image.shape[0:2]
    if col >= 0 and col < width:
        for row in range(0, height):
            pv = image[row][col]
            if(pv[0] > 0 and pv[1] > 0 and pv[2] >0):
                res_row = row
                break
    return res_row

def getEnhancedEdgeImageFromEdgeImage(edge_Image):
    edge_SrcImage = edge_Image
    height, width = edge_SrcImage.shape[0:2]
    for col in range(1, width):
        for row in range(0, height):
            pixel_value = edge_SrcImage[row][col]  # 计算红绿蓝三波段的平均值
            if isPureWhite(pixel_value):
                r_last = getRowNumberSpecificCol(edge_SrcImage, col - 1)
                if r_last:
                    if row > r_last:
                        minR, maxR = r_last, row
                        for k in range(minR, maxR):
                            edge_SrcImage[k][col - 1][0] = 255
                            edge_SrcImage[k][col - 1][1] = 255
                            edge_SrcImage[k][col - 1][2] = 255
                    else:
                        minR, maxR = row, r_last
                        for k in range(minR, maxR):
                            edge_SrcImage[k][col][0] = 255
                            edge_SrcImage[k][col][1] = 255
                            edge_SrcImage[k][col][2] = 255
    # cv2.imshow("Enhanced-edge-image", edge_SrcImage)
    return edge_SrcImage

def getFileExtensionname(filename):
    res = ".png"
    dot_index = -1
    for i in range(len(filename), 0):
        if filename[i] == '.':
            dot_index = i
            break
    if dot_index != -1:
        res = filename[dot_index: len(filename)-1]
    return res

if __name__ == '__main__':
    origin_pic_filename = "D:/test.png"
    sky_ground_threshold = 30
    isDownSampling = False
    if (len(sys.argv) == 1):
        print(sys.argv[0])
        origin_pic_filename = ""
    elif(len(sys.argv) == 2):
        origin_pic_filename = str(sys.argv[1])
    elif(len(sys.argv) == 3):
        origin_pic_filename = str(sys.argv[1])
        sky_ground_threshold = int(sys.argv[2])
    elif (len(sys.argv) == 4):
        origin_pic_filename = str(sys.argv[1])
        sky_ground_threshold = int(sys.argv[2])
        if(int(sys.argv[3]) == 1):
            isDownSampling = True
    if origin_pic_filename != "" and sky_ground_threshold > 0:
        print(("输入图片文件名为:{0}").format(origin_pic_filename))
        print(("天空地面分界灰度阈值为:{0}").format(sky_ground_threshold))
        suffix_name = getFileExtensionname(origin_pic_filename)
        print(("后缀名为:{0}").format(suffix_name))

        srcImage = cv2.imread(origin_pic_filename)
        inputSrcImage = srcImage
        if isDownSampling:
            inputSrcImage = cv2.pyrDown(inputSrcImage)
        height, width = inputSrcImage.shape[0:2]
        print(("高度:{0}, 宽度:{1}").format(height, width))
        cv2.namedWindow('downsampling-image', cv2.WINDOW_AUTOSIZE)
        cv2.imshow("downsampling-image", inputSrcImage)
        Sobelx = cv2.Sobel(inputSrcImage, cv2.CV_64F, 1, 0)
        Sobely = cv2.Sobel(inputSrcImage, cv2.CV_64F, 0, 1)
        Sobelx = cv2.convertScaleAbs(Sobelx)
        Sobely = cv2.convertScaleAbs(Sobely)
        # cv2.imshow("sobel-x-Abs", Sobelx)
        # cv2.imshow("sobel-y-Abs", Sobely)
        Sobelxy = cv2.addWeighted(Sobelx, 0.5, Sobely, 0.5, 0)
        cv2.namedWindow('sobel-xy', cv2.WINDOW_AUTOSIZE)
        cv2.imshow('sobel-xy', Sobelxy)
        edgeImage = getImage(height, width, 3)
        for col in range(0, width):
            for row in range(0, height):
                pixel_value = Sobelxy[row][col]  # 计算红绿蓝三波段的平均值
                if isWhite(pixel_value, sky_ground_threshold):
                    edgeImage[row][col][0] = 255
                    edgeImage[row][col][1] = 255
                    edgeImage[row][col][2] = 255
                    break
        cv2.namedWindow('edge-image', cv2.WINDOW_AUTOSIZE)
        cv2.imshow('edge-image', edgeImage)
        cv2.imwrite(origin_pic_filename.replace(suffix_name, "-ZGetEdge.png"), edgeImage)
        enhanced_edgeImage = getEnhancedEdgeImageFromEdgeImage(edgeImage)
        cv2.namedWindow('enhanced-edge-image', cv2.WINDOW_AUTOSIZE)
        cv2.imshow('enhanced-edge-image', enhanced_edgeImage)
        cv2.imwrite(origin_pic_filename.replace(suffix_name, "-EnhancedEdge.png"), enhanced_edgeImage)

        for col in range(0, width):
            for row in range(0, height):
                pixel_value = enhanced_edgeImage[row][col]  # 计算红绿蓝三波段的平均值
                if isPureWhite(pixel_value):
                    if row+2 < height:
                        inputSrcImage[row+2][col][0] = 0
                        inputSrcImage[row+2][col][1] = 0
                        inputSrcImage[row+2][col][2] = 255
                    else:
                        inputSrcImage[row][col][0] = 0
                        inputSrcImage[row][col][1] = 0
                        inputSrcImage[row][col][2] = 255
                    # inputSrcImage[row][col][0] = 0
                    # inputSrcImage[row][col][1] = 0
                    # inputSrcImage[row][col][2] = 255
                    # break #最开始从每列遍历从上到下找第一个分界点就停止才用break

        cv2.namedWindow('RedEdge-image', cv2.WINDOW_AUTOSIZE)
        cv2.imshow('RedEdge-image', inputSrcImage)
        cv2.imwrite(origin_pic_filename.replace(suffix_name, "-RedEdge.png"), inputSrcImage)
        cv2.waitKey(0)
        cv2.destroyAllWindows()
        print('Success!')
        cv2.waitKey()
        cv2.destroyAllWindows()

3、运行结果

test.jpg下载

3.1 非下采样+边缘检测

cpp 复制代码
python GetSkyLine.py test.jpg  100

|--------------------------------------------------------------------------------------------------------------------------|---------------------------------------------------------------------------------------------------------------------|
| 原始图片 | 边缘点图片 |
| 边缘增强图片 | sobel-xy处理后图片 |
| downloadsampling图片 | 红色边缘叠加图片 |

3.2 下采样+边缘检测

cpp 复制代码
python GetSkyLine.py test.jpg  50  1

原始图片

|--------------------------------------------------------------------------------------------------------------------------|---------------------------------------------------------------------------------------------------------------------|
| 边缘点图片 | 边缘增强图片 |
| downloadsampling图片 | sobel-xy处理后图片 |

红色边缘叠加图片

4、小结

在这个人世间,每个人都是独立的个体,身处浩荡洪流之中,难免身不由己,时而坚定,时而困惑,但我们还是应该永远相信美好的事情终将发生,要心怀一颗感恩的心,相信家人,相信自己,相信未来,坦然面对生活,接受平凡。

关关难过关关过,前路漫漫亦灿灿!

相关推荐
工业3D_大熊6 分钟前
【虚拟仿真】CEETRON SDK在船舶流体与结构仿真中的应用解读
java·python·科技·信息可视化·c#·制造·虚拟现实
SEEONTIME15 分钟前
python-24-一篇文章彻底掌握Python HTTP库Requests
开发语言·python·http·http库requests
Bearnaise15 分钟前
PointMamba: A Simple State Space Model for Point Cloud Analysis——点云论文阅读(10)
论文阅读·笔记·python·深度学习·机器学习·计算机视觉·3d
哇咔咔哇咔1 小时前
【科普】conda、virtualenv, venv分别是什么?它们之间有什么区别?
python·conda·virtualenv
CSXB991 小时前
三十四、Python基础语法(文件操作-上)
开发语言·python·功能测试·测试工具
亚图跨际2 小时前
MATLAB和Python及R潜变量模型和降维
python·matlab·r语言·生物学·潜变量模型
IT古董2 小时前
【机器学习】决定系数(R²:Coefficient of Determination)
人工智能·python·机器学习
德育处主任Pro2 小时前
『Django』APIView基于类的用法
后端·python·django
Star Patrick2 小时前
算法训练(leetcode)二刷第十九天 | *39. 组合总和、*40. 组合总和 II、*131. 分割回文串
python·算法·leetcode
武子康3 小时前
大数据-213 数据挖掘 机器学习理论 - KMeans Python 实现 距离计算函数 质心函数 聚类函数
大数据·人工智能·python·机器学习·数据挖掘·scikit-learn·kmeans