Image Processing using Matlab

Image processing is the process of manipulating the digital properties of an image to improve its quality or to get the desired information from the image. It entails importing the image in the image processing application, analyzing the image, then manipulating the image to get a suitable output that can produce desired results. <!--more--> In this article, we are going to discuss the basics of image processing and analysis using Matlab to determine image characteristics, adjusting image features, and improving image quality.

Prerequisites #

• Have Matlab installed in your computer.
• An understanding of Matlab basics.

Image importation in Matlab #

Importing an image involves bringing the image to the current directory for Matlab. This makes it possible to use the image.

To do this operation, Open Matlab and execute the command below:

i = imread('name of the image;') %This will assign the image to i
imshow(i);

Enhancement of an image using Matlab #

This is the process of improving digital images to get more suitable results for visual display. Image enhancement can be done through image filtering and deblurring.

Image filtering #

This is a form of image enhancement that emphasizes or omits the selected properties of an image. Image filtering mainly involves altering the concentration of certain pixels in an image.

Color filtering makes an image more attractive or stresses certain image pixels. For instance, green can stress vegetation, while blue emphasizes water bodies.

This process can either make an image reddish, greenish or bluish depending on the concentration level applied. imhist function gives a graphical presentation of color concentration in an image per pixel (Histogram).

i =  imread('nyali.jpg');
imshow(i)
Red = i(:,:,1);
Green = i(:,:,2);
Blue = i(:,:,3);
temp = i;

imhist(Red);

imhist(Green);

imhist(Blue);

figure;
temp = i;
temp(:,:,1) = temp(:,:,1) + 100;
imshow(temp);

figure;
imshow(i)
temp = i;
temp(:,:,2) = temp(:,:,2) + 100;
imshow(temp);

temp = i;
temp(:,:,3) = temp(:,:,3) + 100;
imshow(temp);

Image deblurring #

Deblurring an image increases the clarity of the image by making blurry pixels on the image more visible. To carry out this, we first import the image using the code below:

i =  imread('nyali.jpg');       %this code imports the image
imshow(i)

• Create a blurred image from the original image i. This image will be used as a simulator for the deblurring processes.
• Begin by making a point spread function (PSF) by using the fspecial function with specified linear motion (for my case I will use 50 pixels at an angle of 10 degrees) then convolve the PSF with the image by using imfilter function.

PSF = fspecial('motion',50,10);
Idouble = im2double(i);
blurred = imfilter(Idouble,PSF,'conv','circular');
imshow(blurred)

• Use the command deconvwnr to restore the blurred image.

wnr1 = deconvwnr(blurred,PSF);
imshow(wnr1)

Obtaining the number of objects contained in an image #

The number of objects contained in an image can be determined through the following steps:

• Removal of objects whose numerical value is not required.
• Making the image have a uniform background by removing its original background.
• Changing the image to a grayscale image.
• Creating a binary version of the image will allow for a numerical analysis of the objects to be analyzed.

The following code snippet is used in the process.

i = imread('imageName'); %import the image
imshow(i)

se = strel('disk',150);
background = imopen(i,se);  %Performs morphological openning
imshow(background)

• Remove the background approximation image from the original image. This will form a resulting image with a uniform background but a bit dark.

i2 = i - background; 
imshow(i2)

• Change the image format from RGB to grayscale. The newly processed image will be assigned to i3.

i3 = rgb2gray(i2);
imshowi3

• Use the imbinarize command to create a binary version of the grayscale image i3.

bw = imbinarize(i3);
bw = bwareaopen(bw,50);
imshow(bw)

The binary version of the image enables the performance of object analysis. The accuracy of the results depends on object size, parameters connectivity, and the spacing between the objects.

Note that the green-colored objects are not present in the binary version of the image as true colors(red, green, and blue) can not be binarized.

cc = bwconncomp(bw) % shows image information

Finding an area in an image of a specific color #

Areas of objects with different colors in an image can be determined using Matlab through color thresholding. Thresholding is assigning pixels to certain classes based on intensity.

Color thresholding also enables us to determine areas of selected features in a map. For example, finding the area of a water body in a satellite Map.

To demonstrate this concept, I will determine the area of a body of water from a satellite map picture. The picture has a resolution of 480 * 494 pixels and a depth of 32 as per satellite.

i = imread('l.victoria.PNG'); %import the image
imshow(i)

i2 = rgb2gray(i); % make a grayscale image of i
imshow(i2)

imhist(i2) %histogram for pixel distributions

From the histogram, the X-axis represents the intensity value while the Y-axis represents pixel count.

• Open APPS in Matlab window toolbar, scroll down to image processing and computer vision, then click on the color threshold.
• On the new window load an image from the workspace and then choose a color space, click HSV.

• Adjust the knob labeled H to remove the background colors, adjust S and V to smoothen the boundary of the object to be analyzed.

• Create a binary version of the resultant image by clicking on the show binary button.

• Import the binary version to the workspace for further analysis, it will be labeled as BW. The shape of the binary version resembles the shape of the lake, we can visually compare the two images using the function imshowpair.

imshowpair(i,BW,'montage')

• To find the image statistics of the binary version of the image, use the regionprops function.

stats = regionprops('table',BW,'all')

The area of the region in pixels is the sum of the area props in the table.

areainpixels = sum(props.Area)

The area can be converted into square kilometers using a scale that may be provided on the map. Usually, the scale has a predetermined value indicated in meters, miles, or kilometers.

Use the imtool(i) function to measure the distance between the beginning and the end of the linear scale. The measured value will be in pixels, and it represents the predetermined dimension of the scale. Convert from pixel to the required unit.

Conclusion #

Matlab provides a perfect environment for image processing, as the commands and snippets are easy to follow and apply.

Image processing has a wide range of application areas, such as;

• Photography.
• Geographical analysis of images for data.
• Decorations.
• Understanding biological structures.
• Machine vision, and entertainment.

The accuracy of information extracted from an image depends on the quality of the tool used to process the image, and Matlab provides better tools for image processing.

Happy coding!

---

Peer Review Contributions by: Mercy Meave