Department of Computing COURSE: Image Processing MARKS: COURSE CODE: BCM2063 SEMESTER: 2 2020-21 /100 TOPIC : Chapter 1-7 TYPE OF ASSESSMENT: ASSIGNMENT ASSESSMENT NO: ASSIGNMENT TIME / DEADLINE: 3 WEEKS STUDENT NAME: SIGNATURE: STUDENT ID: SECTION: FEEDBACK (If Any): LECTURER NAME: Ms. S. HAMEETHA BEGUM SIGNATURE: INSTRUCTIONS TO CANDIDATE 1. Read the following assignment task instruction and you are required to produce the source code and report as per the instruction given. 2. Create your own folder and name it with your id number. Ex: MC14222 3. Submit the softcopy in Moodle LMS course page within the time given. (Late submission will lead to mark deduction). 4. You are required to demonstrate the code. Describe the code and run the program during your demonstration. 5. You should be able to make amendments to the code during the presentation session according to the instruction. 6. All submitted source code will be checked for plagiarism. Assignment Task Instruction Guidelines for Assignment i. ii. iii. iv. This assignment must be developed in a group (consists of max 3 students per group or minimum 1 student per group) using MATLAB software. Each group is allowed to implement only one (1) title of your own. (Refer to titles below) Refer to the rubrics for the assessment criteria(s). Softcopy of your source code and report must be submitted in Moodle LMS and presented to your lecturer Assignment Requirement 1. Must have at least one (1) of the following techniques/elements: a. Intensity Transformation and Spatial Filtering b. Frequency Domain Processing c. Image Restoration and Noise Removal d. Colour Image Processing e. Image Compression f. Image Segmentation 2. Nice Graphical User Interface 3. A Selection Menu (if any) Suggested Assignment Titles: • Car number plate Image Segmentation • Segmentation of flower petals • Image compression for database of Employee /Student photos • Enhancing and restore Medical X-Ray images for better diagnosis • Identifying Edges / Tumor in Brain –MRI -Medical images Assignment Submission 1. Submit your Report (.DOCX) and source code project embedded in .ZIP or .RAR file format (TWO FILES) in Moodle LMS (Refer Appendix 1) 2. Late submission will be given penalty (Marks deduction). Refer to the rubric for assessment in the evaluation form. BCM2063 Page 2 of 4 BCM2063 – Image Processing Rubric for Assessment (**** Should be attached with your report for marking purpose****) Course Code : BCM2063 Course Name : Image Processing Assignment Title : Student Id Numbers Group No CRITERIA LEVEL OF ACHIEVEMENT WEIGHT AGE 1 3 5 less than 40% problem identification and logical proposal related to image processing techniques apply single image input to be tested 60% problem identification and logical proposal related to image processing techniques 3 images are tested in the experiment More than 80% problem identification and logical proposal related to image processing techniques More than 5 images are tested in the experiment 0.5 30% able to discuss and analyze the results 60% able to discuss and analyze the results More than 80% able to discuss and analyze the results 1.5 Presentation of the final output, 30% complete Presentation of the final output, 100% complete 0.8 Calculation without error 1 CO1 Report a propose alternative solutions. Demonstrate appropriate image input relevant to the problem Analyze the experimental results CO2 The presentation of final output 1 Usage of arithmetic expression (calculation) The overall program structure 6 or 7 errors. Presentation of the final output, 60% complete 2 or 3 errors. 6 or 7 errors. 2 or 3 errors. Ability to execute without error 1 correct usage of selection control statement (if/if..else/case) 6 or 7 errors. 2 or 3 errors. The selection control statement without error 1.5 correct usage of loops statement (for/dowhile/while...etc) 6 or 7 errors. 2 or 3 errors. The loops control statement without error 1.5 SCORE GAINED MARK able to process array data (image) correctly 6 or 7 errors. 2 or 3 errors. The programme is working properly without any error 1.5 Use correct calling function 30% correct usage of sequential statement 60% correct usage of sequential statement 100% correct usage of sequential statement 1.2 Use appropriate process for each function body 6 or 7 errors. 2 or 3 errors. The programme is working properly without any error 1.5 able to show different types of quality image measurements 1 parameter in quality image measurements 3 parameters in quality image measurement s More than 5 parameters in quality image measurements 2 Less than 40% assignment organization complete Less than 60% assignment organization complete 6 logical references in the report More than 80% assignment organization complete 1 At least 10 logical references in the report 0.5 CO3 LL1: Structure Report LL1: (References) LL2: Learning 2 logical references in the report Self Limited ability to implement knowledge Able to to implement knowledge with assistance from lecturer Able to implement knowledge to a given problem independenly 1 LL3:Effort in completing given tasks (experimental results) Less effort to complete task 30% complete Sufficient effort to complete task 60% complete Excellent effort to complete task, 100% complete 0.5 Grand Total 90 Demonstration 10 Net Total 100 Overall Comments: BCM2063 Page 4 of 4 Table of Contents 1.0 INTRODUCTION .......................................................................................... 2 2.0 OBJECTIVES ..............................................................................................2-3 3.0 MATLAB CODE .........................................................................................3-9 4.0 TESTED IMAGES (CANCER IMAGES) ..............................................10-11 5.0 RESULTS & DISCUSSION......................................................................... 12 5.1 Segmentation Techniques ............................................................................ 12 5.1.1 Thresholding basedmethod………………………………………………….……12 5.2 HSV COLOR Model………………………………………………………….……12 6.0 RUNNING THE APPLICATION ...........................................................13-19 7.0 CONCLUSION ............................................................................................ 19 8.0 REFERENCES.............................................................................................. 19 1.0INTRODUCTION Nowadays people can manipulate digital photos by using computer to edit, change and improve the image. Images can be obtained using different types of devices, such as: mobile phone and digital computer. Social media programs are one of the most important programs that carry a billion pictures in limited hours. People can share it through their programs and take the photos they need in their daily practices. Cancer is one of the chronic diseases that appeared in the current era and is called "the disease of the age". This disease is more prevalent among age groups, and fortunately, there is a cure for it. Royal Hospital is one of the most important hospitals in the Sultanate that revealed early cancer in the body of a 34-year-old woman. This is done by applying a solution and making the necessary radiation to detect the cancerous cells and knowing where these cells are concentrated. The person, when infected with cancer and visiting the hospital, the name is stored among the infected patients and is stored in the hospital's database. The hospital performs medical statistics that the Ministry of Health undertakes to know the number of the disease in the oncology department, its purpose, in order to avoid and reduce the infection from the disease. The process of storing names in the database is easy to refer to in order to know the infected persons and treat them with chemical solutions to improve their physical condition. Finally, in this project we have used various techniques to correctly identify cancer cells and where these cells are located in the body, including: Logarithm, sharpening, histogram equalization, restoration techniques, mean and median filtering, Salt and pepper noise and segmentation. 2.0 OBJECTIVE The project aims to develop cancerous cells in the body to help the doctor discover the affected cell in the body by developing a graphical user interface using Matlab. Several images are tested to find the position of cancer cells in the body, and this is done by the user's choice of the image and the comparison of results to obtain an accurate result, in addition to the ability to apply to remove any impurities on the images and improve them for the better. There are also several techniques and functions used in this project, such as Image segmentation technology with the aim of dividing the image into several parts and separating the affected cancer cells. The main task is that the application should allow the user to perform image improvement treatment and direct discovery of cancer cells and save them in another file without the need to formally choose the image. 3.0 MATLAB CODE Main program file (project.m) clc; while(1) disp('Enter your choice:'); disp('1 - GrayScale images'); disp('2 - images Detection'); disp('3 - Exit'); choice = input(''); switch choice case 1 dir_path = 'C:\Users\student\Desktop\MATLAB-projects\Project-IP\images\'; dir_path = strcat(dir_path,'\'); dirpath = strcat(dir_path,'*.jpg'); files = dir(dirpath); file_names={files.name}; sf=size(file_names); for i=1:sf(2) fprintf('Image %d is in progress...\n',i); s1 = strcat(dir_path,char(file_names{i})); M = imread(s1); M = rgb2gray(M); s2 = strcat('C:\Users\student\Desktop\MATLAB-projects\Project-IP\GrayScaleimages\',char(file_names{i})); imwrite(M,s2,'JPG'); clc; continue; end case 2 dir_path = 'C:\Users\student\Desktop\MATLAB-projects\Project-IP\images\'; dir_path = strcat(dir_path,'\'); dirpath = strcat(dir_path,'*.jpg'); files = dir(dirpath); file_names={files.name}; sf=size(file_names); for i=1:sf(2) fprintf('Image %d is in progress...\n',i); s1 = strcat(dir_path,char(file_names{i})); M = imread(s1); M = Detection(M); s2 = strcat('C:\Users\student\Desktop\MATLAB-projects\Project-IP\Detectedimages\',char(file_names{i})); imwrite(M,s2,'JPG'); end case 3 break; otherwise clc; uiwait(msgbox('Please Enter correct choice!!!','','modal')); end end clc; disp('Application Closed!'); close all; Select Image (Button) [filename, pathname]= uigetfile({'*.jpg';'*.bmp';'*.png'},'file selector'); g=strcat(pathname,filename); Originalimg =imread(g); axes(handles.axes1); imshow(Originalimg);title('Original image'); Gray Scale Image (Button) Gray= getimage(handles.axes1); Gray= rgb2gray(Gray); axes(handles.axes2); imshow(Gray);title('GrayScale image'); cla(handles.axes5); set(handles.axes5,'visible','off'); cla(handles.axes6); set(handles.axes6,'visible','off'); RGB to HSV Image (Button) HSV= getimage(handles.axes1); hsv = rgb2hsv(HSV); axes(handles.axes3); imshow(hsv);title('HSV image'); axes(handles.axes4); imhist(hsv);title('Histogram of HSV image'); cla(handles.axes5); set(handles.axes5,'visible','off'); cla(handles.axes6); set(handles.axes6,'visible','off'); Cancer cells Segmented and Detected (Button) x= getimage(handles.axes2); x=double(x); tot=0; [a,b]=size(x); y=zeros(a,b); for i=1:a for j=1:b y(i,j)=0; end end for i=1:a for j=1:b tot=tot+x(i,j); end end thr=tot/(a*b); for i=1:a for j=1:b if x(i,j) > thr y(i,j)=0; else y(i,j)=1; end end end axes(handles.axes5); imshow(y), title('Segmented Image'); x= getimage(handles.axes2); C=double(x); for i=1:size(C,1)-2 for j=1:size(C,2)-2 %Sobel mask for x-direction: Gx=((2*C(i+2,j+1)+C(i+2,j)+C(i+2,j+2))(2*C(i,j+1)+C(i,j)+C(i,j+2))); %Sobel mask for y-direction: Gy=((2*C(i+1,j+2)+C(i,j+2)+C(i+2,j+2))(2*C(i+1,j)+C(i,j)+C(i+2,j))); %The gradient of the image x(i,j)=sqrt(Gx.^2+Gy.^2); end end axes(handles.axes6); imshow(x); title('Detected Image'); Clear ALL (Button) cla(handles.axes1,'reset'); cla(handles.axes2,'reset'); cla(handles.axes3,'reset'); cla(handles.axes4,'reset'); cla(handles.axes5,'reset'); cla(handles.axes6,'reset'); Exit (Button) close all; Menu (Intensity transformation and spatial filtering) ❖ Power law transformation delete(findall(findall(gcf,'Type','axe'),'Type','text')); [filename, pathname]= uigetfile({'*.jpg';'*.bmp';'*.png'},'file selector'); g=strcat(pathname,filename); g =imread(g); f=rgb2gray(g); f = im2double(f); [m n]=size(f); c = 2; y=1.2; for i=1:m for j=1:n s(i,j)=c*(f(i,j)^y); end end axes(handles.axes1); imshow(f),title('Original Image'); axes(handles.axes2); imhist(f),title('Histogram of Original Image'); axes(handles.axes3); imshow(s),title('After Power-Law Transformation'); axes(handles.axes4); imhist(s),title('Histogram After Power-Law Transformation'); cla(handles.axes5); set(handles.axes5,'visible','off'); cla(handles.axes6); set(handles.axes6,'visible','off'); ❖ Sharpening using Laplacian delete(findall(findall(gcf,'Type','axe'),'Type','text')); [filename, pathname]= uigetfile({'*.jpg';'*.bmp';'*.png'},'file selector'); g=strcat(pathname,filename); g =imread(g); A=rgb2gray(g); axes(handles.axes1); imshow(A),title('Original Image'); axes(handles.axes2); imhist(A),title('Histogram of Original Image'); I1=A; I=zeros(size(A)); I2=zeros(size(A)); F1=[0 1 0; 1 -4 1; 0 1 0]; A=padarray(A,[1,1]); A=double(A); for i=1:size(A,1)-2 for j=1:size(A,2)-2 I(i,j)=sum(sum(F1.*A(i:i+2,j:j+2))); end end I=uint8(I); axes(handles.axes3); imshow(I);title('Filtered Image'); axes(handles.axes4); imhist(I),title('Histogram of Filtered Image'); B=I1-I; axes(handles.axes5); imshow(B);title('Sharpened Image'); axes(handles.axes6); imhist(B),title('Histogram of Sharpened Image'); ❖ Logarithmic delete(findall(findall(gcf,'Type','axe'),'Type','text')); g=getimage(handles.axes1); img= im2double(g); c=6; [M,N]=size(img); for x = 1:M for y = 1:N m=double(g(x,y)); z(x,y)=c.*log10(1+m); end end axes(handles.axes1); imshow(g),title('Original Image'); axes(handles.axes2); imhist(g),title('Histogram of Original Image'); axes(handles.axes3); imshow(z),title('After Logarithmic'); axes(handles.axes4); imhist(z),title('Histogram After Logarithmic'); cla(handles.axes5); set(handles.axes5,'visible','off'); cla(handles.axes6); set(handles.axes6,'visible','off'); ❖ Histogram equalization delete(findall(findall(gcf,'Type','axe'),'Type','text')); h=getimage(handles.axes1); axes(handles.axes1); imshow(h), title('Original Image'); axes(handles.axes2); imhist(h), title('Histogram of Original Image'); q= histeq(h); axes(handles.axes3); imshow(q), title('Enhanced Image'); axes(handles.axes4); imhist(q), title('Histogram of Enhanced Image'); cla(handles.axes5); set(handles.axes5,'visible','off'); cla(handles.axes6); set(handles.axes6,'visible','off'); Menu (ImageRestoration.m) clc; while(1) disp('Enter your choice:'); disp('1 - Select Image'); disp('2 - Add Noise'); disp('3 - Add Filter'); disp('4 - Exit'); choice = input(''); switch choice case 1 [filename, pathname] = uigetfile({'*.jpg';'*.bmp';'*.png'},'File Selector'); img = strcat(pathname, filename); if isempty(img) clc; continue; end img = imread(img); img_gray = rgb2gray(img); axes(handles.axes1); imshow(img),title('Original Image'); axes(handles.axes2); imshow(img_gray),title('Grayscale Image'); cla(handles.axes3,'reset'); cla(handles.axes4,'reset'); clc; case 2 clc; while(1) img_gray = getimage(handles.axes2); if isempty(img_gray) message = msgbox('Please Select Image!!!'); disp(message); clc; break; end disp('Enter your choice:'); disp('Type of Noise :'); disp('1 - Speckle Noise'); disp('2 - Salt and Pepper Noise'); disp('3 - Back'); type = input(''); switch type case 1 speckle = imnoise(img_gray,'speckle'); axes(handles.axes3); imshow(speckle),title('Speckle Noise'); clc; cla(handles.axes4,'reset'); break; case 2 salt_pepper = imnoise(img_gray,'salt & pepper',0.13); axes(handles.axes3); imshow(salt_pepper),title('Salt and Pepper Noise'); clc; cla(handles.axes4,'reset'); break; case 3 clc; break; otherwise clc; message = msgbox('Please Enter correct choice!'); disp(message); end end case 3 clc; while(1) noise_img = getimage(handles.axes3); if isempty(noise_img) message = msgbox('Please select image and add noise to it!'); disp(message); clc; break; end disp('Enter your choice:'); disp('Filter Technique :'); disp('1 - Average Filter'); disp('2 - Median Filter'); disp('3 - Back'); type = input(''); switch type case 1 I = double(noise_img); for i=1:size(I,1)-2 for j=1:size(I,2)-2 %Calculate Average: Average = round((I(i,j)+I(i+1,j)+I(i+2,j)+I(i,j+1)+I(i+1,j+1)+I(i+2,j+1)+I(i,j+2)+I(i +1,j+2)+I(i+2,j+2))/9); noise_img(i,j) = Average; end end filtered_img = noise_img; axes(handles.axes4); imshow(filtered_img),title('Average Filter'); clc; break; case 2 I = double(noise_img); for i=1:size(I,1)-2 for j=1:size(I,2)-2 %Find Median: Median = [I(i,j) I(i+1,j) I(i+2,j) I(i,j+1) I(i+1,j+1) I(i+2,j+1) I(i,j+2) I(i+1,j+2) I(i+2,j+2)]; Median = sort(Median); Med = Median(5); noise_img(i,j) = Med; end end filtered_img = noise_img; axes(handles.axes4); imshow(filtered_img),title('Median Filter'); clc; break; case 3 clc; break; otherwise clc; message = msgbox('Please Enter correct choice!'); disp(message); end end case 4 project otherwise clc; message = msgbox('Please Enter correct choice!'); disp(message); end end clc; disp('Application Closed!'); close all; Detection function (Detection.m) function x = Detection(x) x= rgb2gray(x); C=double(x); for i=1:size(C,1)-2 for j=1:size(C,2)-2 %Sobel mask for x-direction: Gx=((2*C(i+2,j+1)+C(i+2,j)+C(i+2,j+2))(2*C(i,j+1)+C(i,j)+C(i,j+2))); %Sobel mask for y-direction: Gy=((2*C(i+1,j+2)+C(i,j+2)+C(i+2,j+2))(2*C(i+1,j)+C(i,j)+C(i+2,j))); %The gradient of the image %B(i,j)=abs(Gx)+abs(Gy); x(i,j)=sqrt(Gx.^2+Gy.^2); end end 4.0 TESTED IMAGES (Cancer cells) Type: jpg Type: jpg Type: jpg Type: jpg Type: png Type: jpg Type: jpg Type: jpg Type: jpeg Type: jpg Type: jfif Type: jpg Type: jpg Type: jpg Type: tiff Type: bmp Type: bmp Type: bmp 5.0 Results & Discussion 5.1 Segmentation Techniques : Image segmentation technology is one of the most important techniques used to detect objects and image processes with high accuracy. The process of segmentation is one of the most important techniques in the image processing domain and known as the technique of splitting the images into several parts called segmented. Image segmentation techniques are work properly based on image criteria such as intensity, color, texture, and others. There are different types of image segmentation processing such as (“Region-based method”, “Thresholdingbased method”, “Clustering-based method”, “watershed based-method” and etc )[1]. 5.1.1 Thresholding-based method Threshold methods are one of the simplest ways to segment images into several parts. This method is works based on the intensity level or image intensity properties. It is appropriate with the image that has a black background and bright object to separate the object with different colors and separate backgrounds with other colors like (black and white). 5.2 HSV color model : This model defines the color space in terms of three main components which are Hue, saturation, and value. Hue is a type of color that includes various colors such as red, magenta, blue, cyan, and green or yellow. The range of hue between 0 to 360 degrees. Saturation is the amount of gray level and refers to how much the color is polluted with white colors. The intensity of saturation between 0 to 255 intensities, where 1 refers to white and 0 refers to black. Value considered as the luminance of elements or values[2]. The application of “cancer cells detection Using image processing ” will use the HSV color model to detect the cell who has the disease and make the affected cell clearer. 6.0 Running the Application When running the application, the program will allow the user to select the images from the file by clicking the select image button. After that, the image selected by the user will appear which is called the Original image, the application allows the user to convert the image into gray level intensities and converted the cancer cells to the HSV color model. Furthermore, This application is useful to know the affected cell by using a segmentation technique such as Threshold methods to splitting the affected cell and make clear to the doctor as shown in figure 1. Figure 1 The program shall display the detected cancer cell after the user clicking “cancer cell segmented and detected button”, the program shall allow the user to clear all images and exit from the application as shown in Figure 2. Figure 2 Figure 3 As shown in figure3 the program shall allow the user to select a choice by using the switch statement, the user can enter 1 to select the image, the user can enter 2 to add noise. There are two types of noise are used in this project which are Speckle noise, salt, and pepper noise. Furthermore, the application shall allow the user to exit from the application as shown in figure 3. When the user enters 2, the application shall allow the user to select three choices as shown in figure 4. Figure 4 Figure 5 As shown in figure 5, the application shall allow the user to enter 1 to apply average filtering which is used to remove noise from the image. In addition, the application shall allow the user to enter 2 to apply the median filter which is used to remove noise from an image and enter 4 to back to the previous page. Figure 6 Figure 7 This project will use image restoration techniques for restoring the image that has been degraded by various reasons, For example, Noise, pixel value errors, out of focus blurring or camera motion blurring. (attempts to restore the image that has degraded for many reasons like motion blur, noise and camera misfocus).As shown in figure 6 and figure 7. Figure 8 Figure 9 When the user clicks the advanced button, the application shall allow the user to enter the choice. If the user enters 1, all the images which are stored in a folder called “images” will be converted to grayscale images in a folder called “GrayScale-images” as shown in figure9. Figure 10 Figure 11 When the user clicks the advanced button, the application shall allow the user to enter the choice. If the user enters 2, all the images which are stored in a folder called “images” will be Detected by using image detection techniques in a folder called “Detected-images” images as shown in figure11. Figure 12 The program shall display error messages as shown in Figure 12 whenever the user enters incorrect choice. Figure 13 When the user clicks the advanced button, the application shall allow the user to enter the choice. If the user enters 1, the application will be closed. 7.0 Conclusion: Finally in this project, after developing a GUI application to find the best techniques for accurately identifying cancerous cells in the body. We have learned that each of these techniques has the ability to know the location of cancer cells and know their greater focus to reduce peak and treatment in the body early. Doctors can use these techniques to quickly identify and treat a patient's condition and avoid medical errors by eliminating cancer cells finally. 8.0 References [1], “IMAGE SEGMENTATION TECHNIQUES”, Available on : https://shodhganga.inflibnet.ac.in/bitstream/10603/50846/10/10_chapter%203.pdf [Accessed January. 3, 2020]. [2], “The HSV Color Model in Graphical Design ” Available on https://www.lifewire.com/what-is-hsv-in-design-1078068[Accessed January. 3, 2020]. Table of Contents 1.0 INTRODUCTION ............................................................................................... 2 2.0 OBJECTIVES .................................................................................................. 2-3 3.0 MATLAB CODE ........................................................................................... 3-12 4.0 TESTED IMAGES (STUDENTS PHOTOGRAPHS) ...................................... 12 5.0 RESULTS & DISCUSSION.............................................................................. 13 5.1 COMPRESSION ALGORITHMS USED .............................................. 13 5.1.1 DCT ............................................................................................ 13 5.1.2 LBG ............................................................................................ 13 5.2 RUNNING THE APPLICATION ..................................................... 14-16 5.3 COMPRESSION TECHNIQUES (RESULTS & COMPARISON) 16-18 6.0 CONCLUSION .................................................................................................. 18 7.0 REFERENCES............................................................................................. 19-20 Page | 1 1.0 INTRODUCTION Nowadays, the digital image can be processed by a digital computer, where we are able to edit, change and improve the quality of an image [1]. The first process to get the image in the image processing procedure is called as image acquisition, where we obtain the image using a different type of camera devices such as mobile phones, digital cameras, computers, etc. In social media applications such as Facebook, Instagram, Twitter, WhatsApp, etc. there are many images that are uploaded every day. According to the statistic that in 2014 there are 2088 digital images are uploaded every second. That’s 1.8 billion digital images every day and 657 billion digital images per year [2]. Also, in many of the organization such as universities, colleges, schools, police, etc. a large number of digital images are stored in the database. In the colleges every new year there are new students are joining college and their information needs to be stored in the college database including a photograph of each student. Therefore, they will be a large number of digital images that consume a large space from storage. One of the most important things for the college is to store these photographs in less space where it can be used in the future to recognize students' faces from the photographs. In this project, an application shall be developed to find the best compression technique to store Muscat College students’ photographs in less storage. There are different image compression techniques used in different applications to store a large number of digital images in less storage such as Huffman code, run-length code, Discrete Cosine Transform DCT, Linde Buzo Gray LBG, etc. some type of compression techniques called lossless compression which maintain the quality of the images and some type of compression techniques called lossy compression which will not maintain the quality of the images but will save more storage. 2.0 OBJECTIVE The main objective of this project is to develop an application to find the best compression technique to store Muscat College students' photographs in less storage. MATLAB software will be used to develop a Graphical User Interface GUI application and implement two image compression techniques which are lossless compression using the DCT algorithm and lossy compression using the LBG algorithm. Page | 2 The application shall allow the user to select and test a sample image using both these techniques for any student image he\she selects in order to compare the results and find the most suitable compression technique. Also, the application shall show the size of images before and after applying the compression process and show the compression ratio and relative data redundancy of compressed image/images. The main functionality is that the application shall allow the user to do bulk processing to apply image enhancement and image compression technique to enhance and compress all the photographs of students and store them in less space. 3.0 MATLAB CODE Main program file (Project.m) clc; global OriginalImageSize; global CompressedImageSize; while(1) dlgtitle = 'Enter your choice:'; prompt = sprintf("1 - Select Image\n2 - Low Compression Test\n3 - High Compression Test\n4 - Compress Students Photographs\n5 - Exit"); dims = [1 50]; value = inputdlg(prompt,dlgtitle,dims); choice = str2double(value); switch choice case 1 [filename, pathname] = uigetfile({'*.jpg';'*.bmp';'*.png'},'File Selector'); file = strcat(pathname, filename); if isempty(file) clc; continue; end img = imread(file); axes(handles.axes1); imshow(img); OriginalImageSize = CalSize(file); OriginalImageSize = round(OriginalImageSize,1); set(handles.text1,'string',"Image size: "+OriginalImageSize+" KB"); axes(handles.axes2); imhist(img); cla(handles.axes3,'reset'); cla(handles.axes4,'reset'); cla(handles.axes5,'reset'); cla(handles.axes6,'reset'); set(handles.text2,'string',""); set(handles.text6,'string',""); set(handles.text3,'string',""); set(handles.text7,'string',""); set(handles.text23,'string',""); set(handles.text24,'string',""); Page | 3 case 2 img = getimage(handles.axes1); if isempty(img) uiwait(msgbox('Please Select Image!!!','','modal')); continue; end img = imresize(img,[1048 1048]); compressedImage = DCT(img); compressedImage = imresize(compressedImage,[264 264]); axes(handles.axes3); imshow(compressedImage); imwrite(compressedImage,'Compressed.jpg'); CompressedImageSize = CalSize('Compressed.jpg'); CompressedImageSize = round(CompressedImageSize,1); set(handles.text2,'string',"Image size: "+CompressedImageSize+" KB"); CR = OriginalImageSize/CompressedImageSize; CR = round(CR,2); set(handles.text6,'string',"Compression Ratio: "+CR); RD = 1 - (1/CR); RD = RD * 100; RD = round(RD,2); set(handles.text23,'string',"Relative Data Redundancy: "+RD+"%"); axes(handles.axes4); img = imread('Compressed.jpg'); imhist(img); case 3 img = getimage(handles.axes1); if isempty(img) uiwait(msgbox('Please Select Image!!!','','modal')); continue; end img = imresize(img,[1048 1048]); compressedImage = LBG(img); compressedImage = imresize(compressedImage,[264 264]); axes(handles.axes5); imshow(compressedImage); imwrite(compressedImage,'Compressed.jpg'); CompressedImageSize = CalSize('Compressed.jpg'); CompressedImageSize = round(CompressedImageSize,1); set(handles.text3,'string',"Image size: "+CompressedImageSize+" KB"); CR = OriginalImageSize/CompressedImageSize; CR = round(CR,2); set(handles.text7,'string',"Compression Ratio: "+CR); RD = 1 - (1/CR); RD = RD * 100; RD = round(RD,2); set(handles.text24,'string',"Relative Data Redundancy: "+RD+"%"); axes(handles.axes6); img = imread('Compressed.jpg'); imhist(img); case 4 while(1) dlgtitle = 'Images Type:'; Page | 4 prompt = sprintf("1 - JPG\n2 - BMP\n3 - PNG\n4 - TIFF\n5 Cancel Progress"); dims = [1 50]; value = inputdlg(prompt,dlgtitle,dims); c = str2double(value); if c ==1 type = 'JPG'; break; elseif c ==2 type = 'BMP'; break; elseif c ==3 type = 'PNG'; break; elseif c ==4 type = 'TIFF'; break; elseif c ==5 break; else uiwait(msgbox('Please Enter correct choice!!','','modal')); continue; end end if c == 5 continue; end while(1) dlgtitle = 'Compression Type:'; prompt = sprintf("1 - Low Compression\n2 - High Compression\n3 - Cancel Progress"); dims = [1 50]; value = inputdlg(prompt,dlgtitle,dims); t = str2double(value); if t ==1 break; elseif t ==2 break; elseif t ==3 break; else uiwait(msgbox('Please Enter correct choice!!','','modal')); continue; end end if t == 3 continue; end dir_path = 'C:\Users\moaya\Documents\MATLAB\Workspace\Project\Students Images\'; dir_path = strcat(dir_path,'\'); dirpath = strcat(dir_path,['*.' type]); files = dir(dirpath); Page | 5 file_names={files.name}; sf=size(file_names); OriginalImage= 0; CompressedImage = 0; for i=1:sf(2) fprintf('Image %d is in progress...\n',i); s1 = strcat(dir_path,char(file_names{i})); M = imread(s1); M = Laplacian_sharpening(M); M = Filtering(M); if t==1 M = imresize(M,[1048 1048]); M = DCT(M); M = imresize(M,[264 264]); else M = imresize(M,[1048 1048]); M = LBG(M); M = imresize(M,[264 264]); M = pwr_law_transformation(M); M = logarithmic(M); end BW = im2bw(M,0.6); s2 = strcat(['C:\Users\moaya\Documents\MATLAB\Workspace\Project\Compressed Images\'],char(file_names{i})); imwrite(M,s2,type); s3 = strcat(['C:\Users\moaya\Documents\MATLAB\Workspace\Project\Compressed Black and White Images\'],char(file_names{i})); imwrite(BW,s3,type); OriginalImage = OriginalImage + CalSize(s1); CompressedImage = CompressedImage + CalSize(s2); end if OriginalImage ~= 0 CR = OriginalImage / CompressedImage; OriginalImage = round(OriginalImage,1); CompressedImage = round(CompressedImage,1); CR = round(CR,2); RD = 1 - (1/CR); RD = RD * 100; RD = round(RD,2); disp('Images has been compressed successfuly.'); fprintf('\nTotal size of the original images is: %.1f KB',OriginalImage); fprintf('\nTotal size of the compressed images is: %.1f KB\n',CompressedImage); fprintf('Compression Ratio: %.2f',CR); fprintf('\nRelative Data Redundancy: %.2f%%\n\n',RD); else uiwait(msgbox(['There are no ' type ' images in this folder!!'],'','modal')); end case 5 break; otherwise Page | 6 uiwait(msgbox('Please Enter correct choice!!','','modal')); end end disp('The Application is Closed!!'); close all; DCT function: function L = DCT(img) I1 = img; R = img(:,:,1); I = im2double(R); T = dctmtx(8); B = blkproc(I,[8 8],'P1*x*P2',T,T'); mask = [1 1 1 1 0 0 0 0 1 1 1 0 0 0 0 0 1 1 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]; B2 = blkproc(B,[8 8],'P1.*x',mask); I2 = blkproc(B2,[8 8],'P1*x*P2',T',T); G = img(:,:,2); I = im2double(G); T = dctmtx(8); B = blkproc(I,[8 8],'P1*x*P2',T,T'); mask = [1 1 1 1 0 0 0 0 1 1 1 0 0 0 0 0 1 1 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]; B2 = blkproc(B,[8 8],'P1.*x',mask); I3 = blkproc(B2,[8 8],'P1*x*P2',T',T); B = img(:,:,3); I = im2double(B); T = dctmtx(8); B = blkproc(I,[8 8],'P1*x*P2',T,T'); mask = [1 1 1 1 0 0 0 0 1 1 1 0 0 0 0 0 1 1 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]; B2 = blkproc(B,[8 8],'P1.*x',mask); I4 = blkproc(B2,[8 8],'P1*x*P2',T',T); L(:,:,:) = cat(3,I2, I3, I4); end Page | 7 LBG function (LBG.m) function out = LBG(A) A = rgb2gray(A); [row, col]=size(A); N=row*col; p=8; n=p*p; Nb=N/n; Nc=64; tSet=zeros(Nb,n); indexCM=1; for i= 1 : p : row for j=1 : p : col temp=blocImage(i,j,A(:,:),n,p); for l = 1 : n tSet(indexCM,l)=temp(1,l); end indexCM=indexCM+1; end end temp=randi(Nb,1,Nc); temp=double(temp); codebook=zeros(Nc,n); for i = 1 : Nc for j=1 : n codebook(i,j)=tSet( temp(1,i), j); end end x=double(1); distort_fact=1 ; %tSet iterations = 0; while x > 0.0005 iterations = iterations+1; distort_fact_prev=distort_fact; %%creating Nc classes %%Nb rows with n elements and Nc channels container_cb=zeros(Nc,n); %Maintains the upper filled limit of the container_class count_container=ones(1,Nc); %This stores the indices of the closest codevector present in codebook with %a particular training vector indexClosestMatch=zeros(1,Nb); for i= 1 : Nb index = ClosestMatch( tSet(i,:) , codebook(: , : ) ); indexClosestMatch(1,i)=index; container_cb( index ,: ) =container_cb( index ,: ) + tSet(i,:); count_container(1,index) = count_container(1,index) + 1; end %This updates the codebook using the container_cb codebook=updateCodeBook(container_cb(:,:,:) , codebook(:,:) , count_container(:,:)); %This function updates the value of distortion factor Page | 8 distort_fact=distortion(codebook(:,:) , indexClosestMatch(:,:) , tSet(:,:) ) ; %This calculates the converence of distortion factor x=abs(distort_fact_prev - distort_fact); x=x/distort_fact_prev; end %%creates the compressed image using codebook and indexClosestMatch out = createCompressedImage(indexClosestMatch,codebook,A,p); update code book function (updateCodeBook.m) function y = updateCodeBook(container,cb,count) [Nc, n]=size(container); for i= 1: Nc temp=zeros(1,n); temp(1,:)=container(i,:); temp = temp/count(1,i); if all(temp == 0) else cb(i,:)=temp(1,:); end end y=cb(:,:); end get closest match training vectors function (ClosestMatch.m) function y = ClosestMatch(TV,cb) [Nc, n] = size ( cb ) ; index = 0 ; dist = 2 ^ 32 ; for i=1 : Nc tempdist = 0 ; for j= 1 : n tempdist = tempdist + ( TV( 1 , j ) - cb( i , j ) ) ^ 2 ; end if dist > tempdist dist = tempdist ; index = i ; end end y = index ; end distortion function (distortion.m) function y = distortion(cb,indexClosestMatch,tset) [Nb, n]=size(tset); sum=0; for i=1:Nb for j=1:n sum = sum + ( cb( indexClosestMatch( 1, i ), j ) - tset( i, j ) )^2; end end sol = sum/(Nb*n); y = sol; end Page | 9 obtain training vector function (blocImage.m) function y = blocImage( i, j, A, n, p) y = zeros(1, n); k = 1; for a = i : p+i-1 for b = j : j+p-1 y( 1, k ) = A( a, b ); k = k+1; end end end create compressed image function (createCompressedImage.m) function out = createCompressedImage(indexClosestMatch,codebook,A,p) [row ,col]=size(A); B = zeros (row,col) ; indexCM=1; for i=1 : p : row for j=1 : p :col index=indexClosestMatch(1,indexCM); l=1; for k=0:p-1 for t=0:p-1 B(i+k,j+t)=codebook(index,l); l = l + 1; end end indexCM=indexCM+1; end end out=uint8(B); imwrite(B,'Compressed.jpg') end calculate image size function (CalSize.m) function out = createCompressedImage(indexClosestMatch,codebook,A,p) [row ,col]=size(A); B = zeros (row,col) ; indexCM=1; for i=1 : p : row for j=1 : p :col index=indexClosestMatch(1,indexCM); l=1; for k=0:p-1 for t=0:p-1 B(i+k,j+t)=codebook(index,l); l = l + 1; end end indexCM=indexCM+1; end end out=uint8(B); imwrite(B,'Compressed.jpg') Page | 10 end average filter function (Filtering.m) function f = Filtering(img) f = img; I = double(f); for i=1:size(I,1)-2 for j=1:size(I,2)-2 Average = round((I(i,j)+I(i+1,j)+I(i+2,j)+I(i,j+1)+I(i+1,j+1)+I(i+2,j+1)+I(i,j+2)+I(i+1 ,j+2)+I(i+2,j+2))/9); f(i,j) = Average; end end end function to obtain sharpened image using Laplacian filter (Laplacian_sharpening.m) function sh = Laplacian_sharpening(img) A=img; I1=A; I=zeros(size(A)); F1=[0 1 0; 1 -4 1; 0 1 0]; A=padarray(A,[1,1]); A=double(A); for i=1:size(A,1)-2 for j=1:size(A,2)-2 I(i,j)=sum(sum(F1.*A(i:i+2,j:j+2))); end end I=uint8(I); sh = I1 - I; end logarithmic function (logarithmic.m) function l = logarithmic(g) g = im2double(g); c=3; [M,N]=size(g); for x = 1:M for y = 1:N m=double(g(x,y)); l(x,y)=c.*log10(1+m); end end end power law transformation function (pwr_law_transformation.m) function t = pwr_law_transformation(f) f = im2double(f); [m n]=size(f); c = 1; y=1.2; Page | 11 for i=1:m for j=1:n t(i,j)=c*(f(i,j)^y); end end end 4.0 TESTED IMAGES (STUDENTS PHOTOGRAPHS) Type: jpg Dimensions: 277 x 231 Size: 31.2 KB Type: jpg Dimensions: 348 x 321 Size: 40.7 KB Type: jpg Dimensions: 595 x 576 Size: 48.2 KB Type: jpg Dimensions: 329 x 286 Size: 33.9 KB Type: jpg Type: jpg Type: jpg Dimensions: 316 x 223 Dimensions: 465 x 366 Dimensions: 853 x 610 Size: 31.0 KB Size: 44.1 KB Size: 77.8 KB Type: jpg Type: jpg Type: jpg Type: jpg Dimensions: 1538 x 1228 Dimensions: 496 x 436 Dimensions: 383 x 295 Dimensions: 595 x 688 Size: 42.8 KB Size: 58.2 KB Size: 716 KB Size: 99.4 KB Type: jpg Dimensions: 288 x 239 Size: 36.1 KB Type: jpg Dimensions: 292 x 238 Size: 36.2 KB Type: jpg Dimensions: 329 x 271 Size: 33.7 KB Type: jpg Dimensions: 396 x 426 Size: 47.1 KB Page | 12 5.0 Results & Discussion 5.1. Compression Algorithms Used 5.1.1. DCT The full form of DCT is Discrete Cosine Transform. DCT is a lossless compression technique applies a transformation to the image by converts the spatial domain into the frequency domain [3]. The algorithm program of DCT works by reading the image and take the red color value of the image then convert it to a double image matrix. Then Compute the two-dimensional DCT of 8by-8 blocks in the image using the function dctmtx which returns the N-by-N DCT transform matrix [4]. After that, it will apply DCT on each block and design a filter to remove the coefficients in a zig-zag manner [7]. The same procedure shall be applied to the green color value and blue color value and then combine the output of the three-color components in one image using the cat function. 5.1.2. LBG The full form of LBG is Linde Buzo Gray. LBG is a lossy compression technique that proposes a vector quantization algorithm based on a training sequence [5]. LBG is the most popular approach for obtaining a vector quantizer codebook. The codebook is a source input group in some vector to find the closest value with codebook values. The set of quantizer output points is called a codebook of the quantizer. LBG is based on a k-means algorithm that applies a clustering procedure to a large set of output vector called a training set. K-means set an initial set of k representative patterns. This representative pattern is updated by computing the centroid of training set vectors. In LBG the distortion from one iteration to the next will not increase [6]. Page | 13 5.2. Running the Application When running the application, the program will keep on asking the user to enter his/her choice initially and after the finish from each operation until the user chooses the exit option to close the application as shown in Figure 1. Figure 1 The program shall display error messages as shown in Figure 2 whenever the user enters incorrect choice. Figure 2 Users can choose to select a sample image as shown in Figure 3 in order to apply low compression (Lossless) and high compression (Lossy) test to the selected image. Figure 3 Page | 14 After selecting an image and applying both compression methods, the program will display the compressed images along with the size of images, compression ratio, relative data redundancy and histogram as shown in Figure 4. Now the user can compare the results and decide what is the best compression type to be applied for the Muscat College students’ photographs. Figure 4 After decide which is the best compression method and choose the fourth option (Compress Students Photographs), the program shall ask to select the type of images to be compressed and then choose the compression type as shown in Figure 5 and Figure 6 in order to do bulk processing to enhance and compress all photographs of the students. Figure 5 Figure 6 When user select compression type, the program will compress and enhance all the images which are stored in a folder called “Students Images” Figure 7 and then store the compressed images in a folder called “Compressed Images”. Then will convert the compressed images into black and white images and store them in folder called “Compressed Black and White Images”. Page | 15 Figure 7 All compression progress will be shown in the command window as shown in Figure 8 and Figure 9 to provide information to the user about which image is in compression progress in the current time and how many images are compressed. Also, the program will display a success message after finishing the progress and show the size of the images before and after compression along with compression ratio and relative data redundancy. Figure 8 Low Compression Results Figure 9 High Compression Results 5.3. Compression Techniques (Results & Comparison) Figure 10 shows the output of the compressed image using a low compression method (Lossless). The images are compressed using the DCT algorithm, enhanced using the average filter and resized to 524 x 524 x 3. Page | 16 Figure 10 Figure 11 shows the output of the compressed image using a high compression method (Lossy). The images are compressed using the LBG algorithm, enhanced using the average filter, logarithmic, power law transformation and resized to 264 x 264. Figure 11 Figure 12 shows the output of the compressed black and white image after applying low or high compression method. The compressed images are converted into black and white. Figure 12 In comparison, the lossless compression algorithm (DCT) has saved the color, quality, and information of the images after compression as shown in Figure 10. On the other hand, the lossy compression algorithm (LBG) has achieved a higher compression ratio and higher relative data Page | 17 redundancy percentage and store the images in less storage as compare to lossless compression as shown in Table 1 below. Table 1 Lossless & Lossy Compression results Compression Type Lossless (DCT) Lossy (LBG) Original Images Size 1377.5 KB 1377.5 KB Compressed Images Size 106.9 KB 82.8 KB 12.89 16.63 92.24% 93.99% Result Type Compression Ratio Relative Data Redundancy Lossy compression has saved the information of the images as shown in Figure 11. Although Lossy compression did not save the color like lossless compression. However, Lossy compression is the most appropriate compression technique to save Muscat College students’ photographs in less storage, because the main goal is to save the images in less storage where we can recognize the faces in the future, but the color and the quality is not much important. 6.0 Conclusion All in all, after developing the GUI application to find the best compression technique to store Muscat College students' photographs in less storage, I concluded that: • In lossless compression, only the two-dimensional matrix M x N is affected but in lossy compression, the three-dimensional matrix M x N x 3 is affected. • Lossless compression using the DCT algorithm is suitable to compress the image without quality loss. • The compression techniques are very important to reduce utilized space in memory in order to store more information. • Lossy compression using LBG is not suitable to save the color and quality of the images. But is suitable to save the images in less space as compared to lossless compression. • Lossy compression using LBG is suitable for high-resolution images only, therefore resizing the low-resolution image to the high resolution before compression is most. • Lossy compression is good to achieve a high compression ratio. Page | 18 7.0 References [1] Sachin Seth, “What are the advantages of digital image processing”, www.quora.com, Mar. 22, 2017. [online]. Available: https://www.quora.com/What-are-the-advantages-of-digitalimage-processing. [Accessed Oct. 19, 2019]. [2] Mary Meeker, “Internet Trends 2018”, www.kleinerperkin.com, May. 30, 2018. [online]. Available: https://www.kleinerperkins.com/files/INTERNET_TRENDS_REPORT_2018.pdf. [Accessed January. 2, 2020]. [3] Shrenik Jain, “(DCT) Discrete Cosine Transform in image processing”, www.youtube.com, December. 13, 2017. [online]. Available: https://youtu.be/tW3Hc0Wrgl0. [Accessed January. 3, 2020]. [4] MathWorks, “Discrete Cosine Transform”, www.mathworks.com. [online]. Available: https://www.mathworks.com/help/images/discrete-cosine-transform.html. [Accessed January. 3, 2020]. [5] MathWorks, “IMAGE COMPRESSION USING LBG ALGORITHM”, www.mathworks.com. [online]. Available: https://www.mathworks.com/matlabcentral/fileexchange/10457-image-compression-using-lbgalgorithm. [Accessed January. 3, 2020]. [6] ITechnica, “Linde Buzo Gray (LBG) algorithm step by step with example”, www.youtube.com, May. 1, 2018. [online]. Available: https://youtu.be/_S-5M8Zf6hc. [Accessed January. 3, 2020]. [7] Dr.K.Somasundaram & Dr.T.Kalaiselvi, “Image Compression using Discrete Cosine Transform(DCT) using MATLAB blockproc() Function”, www.researchgate.net, February. 23, 2015. [online]. Available: https://www.researchgate.net/publication/272676573_Image_Compression_using_Discrete_Cosi ne_TransformDCT_using_MATLAB_blockproc_Function. [Accessed January. 3, 2020]. Page | 19 [8] MathWorks, “How can I display image size, dimensions and color type in the GUI?”, www.mathworks.com, Aug. 18, 2014. [online]. Available: https://www.mathworks.com/matlabcentral/answers/151331-how-can-i-display-image-sizedimensions-and-color-type-in-the-gui. [Accessed December. 12, 2019]. [9] MathWorks, “msgbox”, www.mathworks.com. [online]. Available: https://www.mathworks.com/help/matlab/ref/msgbox.html. [Accessed December. 12, 2019]. [10] MathWorks, “imresize”, www.mathworks.com. [online]. Available: https://www.mathworks.com/help/images/ref/imresize.html. [Accessed December. 12, 2019]. [11] Ulmer Louis, “Image-Compression-Using-Vector-Quantization-with-LBG-algorithm”, getlab.insa-rouen.fr. [online]. Available: https://gitlab.insarouen.fr/lulmer/ASI4_LouisULMER/tree/2f2582aeb2d10a136c3259e5e26513416005b28e/TI/pr ojet_quantification/partie_vectorielle. [Accessed January. 2, 2020]. [12] MathWorks, “How to round a result to two decimal places”, www.mathworks.com, Oct. 24, 2016. [online]. Available: https://www.mathworks.com/matlabcentral/answers/308927-how-toround-a-result-to-two-decimal-places. [Accessed January. 3, 2020]. [13] MathWorks, “How do I print a "%" when using fprintf?”, www.mathworks.com, Mars. 26, 2019. [online]. Available: https://www.mathworks.com/matlabcentral/answers/452606-how-doi-print-a-when-using-fprintf. [Accessed January. 3, 2020]. [14] MathWorks, “im2bw”, www.mathworks.com. [online]. Available: https://www.mathworks.com/help/images/ref/im2bw.html. [Accessed January. 5, 2020]. Page | 20