【图像处理】删除ROI区域之外的噪点并获取外接矩形

只搬烫手的砖

已于 2022-08-07 20:29:24 修改

阅读量779

点赞数

文章标签：图像处理 opencv 计算机视觉

于 2022-08-07 10:09:00 首次发布

本文链接：https://blog.csdn.net/qq_44747572/article/details/126206995

版权

目标：去掉ROI区域以外的噪声，进而获取不规则ROI区域的外接矩形。

其中离散点的去除，采用的是opencv中的高斯滤波+OSTU方法进行滤除，但是该方法仍然会降ROI区域中的某些阈值较低的信息给滤除掉，暂时没有想到更好的办法去除离散点

# Author:yx
# Time:2022.7.21
# Function: 去掉ROI区域以外的噪声，进而获取不规则ROI区域的外接矩形。

from tkinter import Image
import cv2
import numpy as np
import skimage

def zuidawaijie(Img):
    Img_H = Img.shape[0]
    Img_W = Img.shape[1]

    Top = 0
    Bottom = 0
    Left = 0
    Right = 0

    # 识别外边框
    # 识别Top的第一个点
    for i in range(Img_H):
        for j in range(Img_W):
            if Img[i, j].all() != 0:
                Top = i
                break
        if Top != 0:
            break

    # 识别Bottom的第一个点
    for i in range(Img_H):
        for j in range(Img_W):
            if Img[Img_H -1 - i, j].all() !=0:
                Bottom = Img_H -1 - i
                break
        if Bottom != 0:
            break

    # 识别Left的第一个点
    for j in range(Img_W):
        for i in range(Img_H):
            if Img[i, j].all() != 0:
                Left = j
                break
        if Left != 0:
            break

    # 识别Right的第一个点
    for j in range(Img_W):
        for i in range(Img_H):
            if Img[i, Img_W -1 - j].all() !=0:
                Right = Img_W - 1 - j
                break
        if Right != 0:
            break

    # top
    for i in range(Left, Right):
        Img[Top, i] = 255
    # bottom
    for i in range(Left, Right):
        Img[Bottom, i] = 255
    # Left
    for i in range(Top, Bottom):
        Img[i, Left] = 255
    # Right
    for i in range(Top, Bottom):
        Img[i, Right] = 255

    cv2.imshow('Img', Img)
    cv2.waitKey()
    cropImg = Img[Top:Bottom, Left:Right]
    return cropImg,Top,Bottom,Left,Right
   



if __name__ == '__main__':
    img = cv2.imread('./calPosition/1328.png', 0)
    blur = cv2.GaussianBlur(img, (3, 3), 0)
    _, binary_img = cv2.threshold(blur, 0, 255, cv2.THRESH_BINARY + cv2.THRESH_OTSU)
    cropbinary_img,Top,Bottom,Left,Right = zuidawaijie(binary_img)
    cropImg = img[Top:Bottom, Left:Right]
    # 上下平移量 负数上移  正数下移
    DUnum = 100
    # 左右平移量 负数左移  正数右移
    LRnum = 100
    img[Top-DUnum:Bottom-DUnum, Left-LRnum:Right-LRnum] = cropImg
    cv2.imshow('img', img)
    cv2.waitKey()