Discover MakerZone

MATLAB and Simulink resources for Arduino, LEGO, and Raspberry Pi

Learn more

Discover what MATLAB® can do for your career.

Opportunities for recent engineering grads.

Apply Today

can any one suggest any corrections or changes for the code witten by me.

Asked by ARUN SAI on 23 May 2013

As iam doing my project on color filtering method for CFA images based on gradient,i have a problem with my code written so guide me if there is any changes in the code.Even though the noise is added to the image ,but it is still showing that there is noise according to the code.

 i = imread('mandi.tif');
 imshow(i);

%demosaic of image

 rgb = demosaic(i,'bggr'); *%raw image is converted into true color image*
 figure,imshow(rgb);
 m1=mean2(rgb);
 SD1=mean2(stdfilt(rgb));
 snr1=(m1/SD1) *%signal to noise ratio calculation of noise free image*
 for i=500:1:1000
    for j=500:1:1000
        rgb(i,j)=imnoise(rgb(i,j),'gaussian',9); *%noise addition*
    end
 end
 figure,imshow(rgb);
 m2=mean2(rgb);
 SD2=mean2(stdfilt(rgb));
 snr2=(m2/SD2) *%signal to noise ratio calculation of noisy image*
 r=rgb;g=rgb;b=rgb;
 r(:,:,2)=0;
 r(:,:,3)=0;  
 g(:,:,1)=0;
 g(:,:,3)=0;
 b(:,:,1)=0;
 b(:,:,2)=0;
 m = size(rgb,1) 
 n = size(rgb,2)
 for i=2:1:m-1
     for j=2:1:n-1
 A1=abs(r(i-1,j-1)-r(i+1,j+1)); *% horizontal internal gradient*
 B1=abs(r(i-1,j+1)-r(i+1,j-1)); *% vertical internal gradient*
    end
 end
 for i=3:1:m-2
    for j=3:1:n-2
      A2=abs(2*g(i,j)-g(i-1,j-2)-g(i+1,j+2)); *% horizontal external gradient*
      B2=abs(2*g(i,j)-g(i-2,j-1)-g(i+2,j+1)); *% vertical external gradient*
    end
 end

% to find existence of edge or influence of the noise TH (enumeration variable) is used for noise that in the up and down or in any point of g(i,j)

 for k=2:2:m-1
     for l=3:2:n-1
        x=[k-1,k+1];
        y=[l-1,l+1];
up=(A1<B1) & (eq(A2,B2)) & eq(max(abs(2*r(k,y)-r(k-1,l)-r(k+1,l))),l-1);    *%up says that g(i,j-1) is the noise*
down=(A1<B1) & (eq(A2,B2)) & eq(max(abs(2*r(k,y)-r(k-1,l)-r(k+1,l))),l+1); *%down says that g(i,j+1) is the noise*
left=(A1>B1) & (eq(A2,B2)) & eq(max(abs(2*r(x,l)-r(k,l-1)-r(k,l+1))),k-1); %left says that g(i-1,j) is the noise
right=(A1>B1) & (eq(A2,B2)) & eq(max(abs(2*r(x,l)-r(k,l-1)-r(k,l+1))),k+1);
*%right says that g(i+1,j) is the noise*
no=(eq(A1,B1)) & (eq(A2,B2)); 
*%no says that there is no noise and also no edge *
level=(A1>B1) & (A2>B2); 
*% level says that there is a edge in vertical direction*
erect=(A1>B1) & (A2<B2);
*% erect says that there is a edge in horizontal direction*
    end
 end
 p=[1 2 3 4 5 6 7]
 TH=[up down left right no level erect]
 p_TH=p(TH==1)
 for i=2:1:m-1
    for j=2:1:n-1

% switches according to the TH value

 switch p_TH 
    case 1
        r(i,j)=(r(i-1,j)+r(i+1,j)+r(i,j+1))/3;
        r(i,j-1)=r(i,j);
    case 2
        r(i,j)=(r(i-1,j)+r(i+1,j)+r(i,j-1))/3;
        r(i,j+1)=r(i,j);
    case 3
        r(i,j)=(r(i+1,j)+r(i,j-1)+r(i,j+1))/3;
        r(i-1,j)=r(i,j);
    case 4
        r(i,j)=(r(i-1,j)+r(i,j-1)+r(i,j+1))/3;
        r(i+1,j)=r(i,j);
    case 5
        r(i,j)=(r(i-1,j)+r(i+1,j)+r(i,j-1)+r(i,j+1))/4;
    case 6
        r(i,j)=(r(i-1,j)+r(i+1,j))/2;
    case 7
        r(i,j)=(r(i,j-1)+r(i,j+1))/2;
 end
    end
 end
 s=r+g+b;
 figure,imshow(s);
 m3=mean2(s);
 SD3=mean2(stdfilt(s));
 snr3=(m3/SD3) *%signal to noise ratio calculation of image after removal of  noise*

3 Comments

José-Luis on 23 May 2013

And why should anyone do your own work? Do you have a specific question?

ARUN SAI on 25 May 2013

suggest me if there is any changes in the code that is in the computing the enumeration variable TH=up,down,left,right,level,erect,no.As iam indication that there is no noise in the image after running the program

ARUN SAI

2 Answers

Answer by Image Analyst on 25 May 2013

Explain why you are doing this to a gray scale image:

%demosaic of image
rgb = demosaic(i,'bggr');

3 Comments

ARUN SAI on 25 May 2013

my project *color filtering method for CFA images based on gradient *is about the process of removing the noise caused during demosaicking or artifacts caused during image acquisition.So in a single sensor digital camera,the sensor can sense only one of the color ie, R or G or B and the remaining colors are predicted through demosaic function.The demosaic function converts raw image into true color image.

Image Analyst on 25 May 2013

OK - I didn't realize mandi was a raw image until I zoomed in on it. But the lack of comments make it difficult to follow what you are doing and don't really make me want to try. Add comments to explain each step.

Nonetheless I tried to run it and it put up two images then waited a long time and put up a third image and then took many many minutes before I killed it. How long is it supposed to take to finish?

For example, why on earth are you adding noise in a loop one pixel at a time?

%noise addition
for i=500:1:1000
  for j=500:1:1000
      rgb(i,j)=imnoise(rgb(i,j),'gaussian',9);
  end
end

Just do it for the whole 501x501 matrix all in one shot. It will be so much faster.

ARUN SAI on 25 May 2013

no,i am just adding noise to part of an image.and if there is any alternative to add noise, that which is faster give me the code.

Image Analyst
Answer by Jan Simon on 26 May 2013
Edited by Jan Simon on 26 May 2013

I suggest (as usual) to omit the brute clearing header:

 clc;
 clear all;
 close all;

Closing all windows might be convenient in some rare situation, but as soon as you will use several different programs with GUIs, such habits will make you live much harder.

Clearing the command line let warning and error messages disappear. Is this useful?

clear all removes all loaded functions from the memory and reloading them from the hard disk wastes a lot of time. But even worse, it delete all debugger breakpoints, and anything, which impedes debugging is a bad programming pattern. Sometimes clear might help to find bugs, because it cleans up formerly defined variables. E.g. if the variable "iO" (i Oh) was defined earlier, then in a script "i0" (i Zero) is defined, but "iO" (i Oh) used due to a typo. But using functions instead of scripts are much better to catch such problems, because then MLint helps finding such problems.

In spite of these drawbacks the brute clearing header are frighteningly wide-spread in the code of Matlab beginners. Who on earth suggests such stuff?!

Concerning:

switch p_TH *% switches according to the TH value*

Both stars are misplaced here. I guess, this was thought for formatting in forum?

10 Comments

Jan Simon on 26 May 2013

@Jose-Luis: Co-author? I'm sure Image Analyst does not want to be responsible for the outcomes of this project. I'd guess, that hard dollars might be more encouraging.

@ARUN SAI: The voluntary contributors of this forum offer assistance for free. Therefore a commanding tone like "rectify those problems" is not welcome here. Your aim to solve your (!) academic project is no reason or excuse to omit a polite style of conversation.

ARUN SAI on 27 May 2013

sorry,but suggest me to overcome this problem.

ARUN SAI on 5 Jul 2013

sir can it be extended for the video file.if yes how to import video of any extension file.

Jan Simon

Contact us