amr1235/cv-project-1

Computer Vision Project 1

In this project we have implemented some Image Processing Techniques implemented from scratch using C++, we have implemented our Techniques in this readme in lenna image for more illustration.

Appendix

Lenna Image

Prerequsites

1. Visual Studio C++ Workload
2. OpenCV
3. CMake

Build

1. Configure project.

N.b use only after cloning or changing CMakeLists.txt.

$ cmake . -B ./build

2. Build the executable

$ cmake --build ./build/

3. The executable will be in the build/Debug/ folder

$ build/Debug/cv-project-1.exe

4. Clean the build folder (Optional)

$ cmake --build ./build/ --target clean

## 1. Adding Noise To Image

Three types of noise are implemented in this project: Uniform, Gaussian, and Salt & Pepper noise

img::Image add_noise(img::Image image , std::string type ,uchar noise_factor)

Parameters

image : the image on which the should be added, it should be of class img::Image which is implemented in this project.

type : the type of noise you want to add, there are three types you must write any one of them "Gaussian", "Uniform" , and "Salt".

noise_factor : It's the ratio of the noise in the image.

Examples

1.1 Uniform Noise

1.2 Gaussian Noise

1.3 Salt & Pepper Noise

## 2. Image Filtering

Three types of noise are implemented in this project: Median, Mean, and Gaussian filters.

img::Image noise_filter(img::Image source,int size, std::string type ,double std=1)

Parameters

image : the image on which the should be added, it should be of calse img::Image

type : the type of noise filter you want to apply, there are three types you must write any one of them "Gaussian", "Mean" , and "Median".

std : It's the standard deviation of the gaussian kernal, it has a default value equals 1, It should be used only when using Gaussian filter unless it will be useless.

Examples

The following examples for filttering the Salt & Pepper noise implemented in this project.

2.1 Median Filter

2.2 Mean Filter

2.3 Gaussian Filter

3- Edge detection

Three types of detection are implemented in this project: Sobel, Prewitt, and Roberts. sobel :

Image sobel(const Image &img, bool dx = true, bool dy = true)

Image prewitt(const Image &img, bool dx = true, bool dy = true)

Image roberts(const Image &img, bool dx = true, bool dy = true)

cvector<double> canny(const Image &img, double threshold1, double threshold2)

Parameters

Image : Image

dx : if you want to detect the horizontal edges

dy : if you want to detect the vertical edges

Examples

3.1 Sobel Detection

3.2 Prewitt Detection

3.3 Roberts Detection

3.4 canny Detection

4- Histogram and Distribution Curve

void Plot_Histogram(Image img,string image_type)

void plot_Distribution_curve(Image img,string image_type)

Parameters

image : the image you want to plot its histogram and distribution curve, it should be of class img::Image which is implemented in this project.

image_type : the type of image you entered "grayscale" or "color"

Results

4.1 Histogram

4.2 Distribution Curve

5. Equalization

equalization has multiple steps of pipelining such as histogram loop, accumulation loop and mapping loop. this function is returning Image type so it can be used in further steps. the whole function is implemented in image.cpp

Parameters:

img: the image which will be equalized. (image type)

Results:

6. Normalization

Normalization is only dependent on few parameters which are min new value, max new value and the image itself. by the given equation -> (newrange * (value - oldmin) / oldrange) + newmin we can get the new value for a pixel We put the equation in utils.cpp file, while the whole normalization function were put in the mentioned image class.

Parameters:

img: the image which will be normalized. (image type) min: minimum percentage as this function is dynamic (not restricted to specific number). (0 - 100) max: max percentage as this function is dynamic (not restricted to specific number). (0 - 100)

Results:

normalized between 0 and 20 percentage

normalized between 80 and 100 percentage

7. local and global thresholding

7.1 global thresholding

Image globalThreshold(Image &img, float threshold, float mx_value, t_types thresholdType) for global Thresholding

img : the image that you want to threshold

threshold : threshold value

mx_value : the value is set in place of pixel

thresholdType : threshold type can be THRESH_BIN if the value is greater than threshold so the maximum is set to the pixel or THRESH_BIN_INV if the value is greater than threshold so 0 is set to the pixel

7.2 local thresholding

Image localThreshold(Image &img, float mx_value, t_types thresholdType) for local Thresholding it uses an adaptive gaussian thresholding algorithm

img : the image that you want to threshold

threshold : threshold value

mx_value : the value is set in place of pixel

thresholdType : threshold type can be

THRESH_BIN if the value is greater than threshold so the maximum is set to the pixel or THRESH_BIN_INV if the value is greater than threshold so 0 is set to the pixel

Results:

8.1- Transformation from color image to gray scale image

img::convert(img::Image source, "bgr", "gray");

Result

8.2-Plot R,G,B Histograms with its Cumulative Curves

void Plot_Histogram(Image img,string image_type)

void Plot_Cumulative(Image img,string image_type)

Results

Red Histogram

Green Histogram

Blue Histogram

Red Cumulative Curve

Green Cumulative Curve

Blue Cumulative Curve

9. low pass and high pass filter

Image PassFilter(Image &img, float freq_threshold,p_types pass_type) to filter the frequency domain

img : the image that you want to filter

freq_threshold : frequency threshold value

pass_type : pass type can be

LOW_PASS_FILTER it removes all the values that is above the frequency threshold , HIGH_PASS_FILTER it removes all the values that is below the frequency threshold

Results

10. hybrid images

Image hybridImages(Image &img1, Image &img2) it takes two images and merge them in one image using the low frequency of one image and the high frequency of the other image

Results