Skip to content

Latest commit

 

History

History
158 lines (138 loc) · 5.95 KB

README.md

File metadata and controls

158 lines (138 loc) · 5.95 KB

Core_Operation

Descrption:

  • know how to select a ROI (region of interest) in an image.
  • know how to do "image padding" to fill the edge of an image.
  • know how to use "image blending" and "birwise operation" to combine two images.

Outline:

  1. Image ROI
  2. Image padding
  3. Image blending
  4. Bitwise operation

1. Image ROI (Full code in python)

- Input image: opencv.png
- Command Line: python image_ROI.py --image opencv.png

- Method 1: Use pixel range to select the region

ROI_image = Input_Image[22:265, 60:311]
NOTE: 
  - notice that the image pixel x, y position in OpenCV is Input_Image[y, x].
  - ROI region x axis: from pixel 60 to 311, y axis: from pixel 22 to 265.

- Method 2: Use cv2.selectROI("input_image") to select the region

r = cv2.selectROI(Input_Image)
ROI_img = image[int(r[1]):int(r[1]+r[3]), int(r[0]):int(r[0]+r[2])]
NOTE: r = (x, y, width, height), we can use these return region values to crop the image.

2. Image padding (Full code in python)

- Input image: padding_source.png
- Command Line: python image_ROI.py --padding_source.png

- Method: Use cv2.copyMakeBorder(Source, top, bottom, left, right, borderType)

void cv::copyMakeBorder	( InputArray src,
                          OutputArray dst,
                          int 	top,
                          int 	bottom,
                          int 	left,
                          int 	right,
                          int 	borderType,
                          const Scalar & value = Scalar() 
)	

* replicate = cv2.copyMakeBorder(..,cv2.BORDER_REPLICATE)
* reflect = cv2.copyMakeBorder(..,cv2.BORDER_REFLECT)
* wrap = cv2.copyMakeBorder(..,,cv2.BORDER_WRAP)
* constant= cv2.copyMakeBorder(..,cv2.BORDER_CONSTANT,value=BLUE)

2. Image blending (Full code in python)

- Input image: 
  * Background image: dog.jpg
  * foreground image: moon.jpg
- Command Line: python image_blending.py --image dog.jpg --ontop moon.jpg

- Method: Use cv2.addWeighted(src_1,alpha,src_2,beta,gamma)

This function calculates the weighted sum of two images.
dst(I) = Saturate(src(I)*alpha + src(I)*beta + gamma)

2. Bitwise operation (Full code in python)

- Input image: 
  * Background image: dog.jpg
  * foreground image: opencv.png
- Command Line: bitwiseOp.py --image dog.jpg --ontop opencv.png

NOTE: two images need to be same shape.

- Function 1: Overlap two images

STEP 1. If two images not in the same shape, use function "cv2.resize()" to resize the image.

cv2.resize(src, (height, width), interpolation)
  - interpolation:
    * INTER_NEAREST - a nearest-neighbor interpolation
    * INTER_LINEAR - a bilinear interpolation (used by default)
    * INTER_AREA - resampling using pixel area relation. 
    * INTER_CUBIC - a bicubic interpolation over 4x4 pixel neighborhood
    * INTER_LANCZOS4 - a Lanczos interpolation over 8x8 pixel neighborhood

STEP 2. Create mask by using grayscale image and threshold.

Convert color(BGR) image to grayscale image. 

Function: cv2.cvtColor(input_image, flag)
  - Flag:
    * cv2.COLOR_BGR2GRAY: BGR to Gray conversion. 
    * cv2.COLOR_BGR2HSV: BGR to HSV conversion.

More detail on Changing Colorspace

Thresholding: If pixel value is greater than a threshold value, it is assigned one value, 
              else it is assigned another value 

Function: ret,thresh1 = cv.threshold(src, threshold, maxval, type)
  - type:
    * cv.THRESH_BINARY
    * cv.THRESH_BINARY_INV
    * cv.THRESH_TRUNC
    * cv.THRESH_TOZERO
    * cv.THRESH_TOZERO_INV

More detail on Image Thresholding

STEP 3. Use function "cv2.bitwise_and()" to get background and foreground area on two images. Then add together.

NOTE: Use function "cv2.bitwise_not(mask)" which inverts every bit to get opposite mask.

* cv2.bitwise_and
* cv2.bitwise_or
* cv2.bitwise_xor

Usage: https://docs.opencv.org/2.4/modules/core/doc/operations_on_arrays.html
Code in bitwiseOp.py

- Function 2: Add foreground image on the top left (or any desire position): Method similar to Funtion 1.

Code

License

This project is licensed under the MIT License - see the LICENSE.md file for details

Reference & Acknowledgments