Color Detection by using Color space basics

Ever wondered how can i detect the color ..rather one color from many colors and highlight the same..?

This post is all about it…a quick glimpse about the color detection thing is here and also the method which I have used here is Color space conversion..So If you are a color Enthusiast then you should study this post…

Let me also tell you the algorithm used here for color detection is also used once more in another program…

So now let us extract the red things from the images and this time we will do it with perfection and let me also tell you that the we are only going to detect the red color here…

If you want to understand the whole point of this program then you need to first understand the basics
of color conversion..You need to understand how the HSV Colorspace works.

In Simple words

H-Hue which means in this channel we want to know “which” is the color which is present

Saturation-Which means how much amount of color is present which means

higher the number means that there is more color and lesser white light….in other words if you see an area which is perfectly red, which means there is more color present and lesser white light.

If you focus a torch on a coloured ball….the area where the ball shines means that more of white light is getting reflected from the ball from that part and hence that area is said to be more diluted by the white light and hence you would see the shine the color is there but there is more shine in the color…

/*ALL the necessary header files*/

 #include"math.h"
 #include"conio.h"
 #include"cv.h"
 #include"highgui.h"
 #include"stdio.h"
int main()
{
int i,j,k;//for iterations
int temp=0;//if we use a temporary var
/*here lets look at the word “heighthsv” …now lets breadk up this word…here height means

height as a regular IplImage Structure has now the addition “hsv” to the word heigh means this
height attribute is for the image which is color converted to the hsv,Similar conventions
for the monochrome image…so you may find the attribute height for the monochrome image to be
heightmono…So i believe it is easy…*/
int heighthsv,widthhsv,stephsv,channelshsv;
int heightmono,widthmono,stepmono,channelsmono;
uchar *datahsv,*datamono;
int sthreshold=140;

So have a look at this variable very closely….

We are going to use this variable to detect the intensity of the color which is red for this time
So if you have gone through the HSV color space tutorial then you must have got the concept
of what is the saturation in a HSV colorspace and if you did not before proceeding with thsi program

you should have a look at this link

http://en.wikipedia.org/wiki/HSV_color_space
This link according to me is the best link which one can ever
suggest for the understanding the color spaces
Now sthreshold means saturation threshold which means we are only allowing the pixels
above a certain value to be realized so the hue values will allow a color and the saturation value
will allow the color which more intense than the particular value…which is said by the sthreshold
Now the what do the hlower and the hupper mean…..

Now let us have a look at an example…so if red is at 0(of course hue value of red is zero..

So we need to allow red which is of a particular intensity and above
so we have selected the range of the hue value which is allowed to be 175-05
(which always means above 175 and below 10)(0-10 and 175-180)
Now an obvious question the hue range is between 0-360 and the red color is 0

so we must have selected the range to be 355-10(which means above 355 and below 10)…because after the values
approach 360 it again become zero to complete the closed circle of the hue here
But if we closely observe an INTEGER data type in which is of 1 byte(8 BITS) can store a max value of
255…so 0-360 is not realizable…to make it realizable OPENCV has compressed the circle to half
which means if originally if we have a color of the hue value 120 in the HSV color space
now to make that color realizable with the help of only one byte which has a max range of 255
the same color which has a value of 120 originally in the HSV colorspace would have a value
of 60 in opencv….which means we have divided all the original values by 2….so the color red
which is in between 0-360…would now be found in between 0-180 which is 360/2..
…and the numbers after 180 would be again zero so i have chosen the range which would detect
red would be 0-175 so the lower cut off value for the red
would be 175 and the upper cut off would be 05

You may continue copying the code from here.

double hlower=175,hupper=5;/*
here hlower is the lower cut off and the hupper is the upper cut off for hue values of red.*/

i=j=k=0;/*initializing the iteraiton variables to be zero*/

IplImage *frame=cvLoadImage(“3d-09-5balls.jpg”,1);

if(frame==NULL ) {
puts("unable to load the frame");exit(0);}
printf("frame loaded");
IplImage *colimgbot = cvCreateImage( cvGetSize(frame), 8, 3 );

IplImage *monoimgbot = cvCreateImage( cvGetSize(frame), 8, 1 );

//——————————————————————

heighthsv = colimgbot->height;
widthhsv = colimgbot->width;
stephsv =colimgbot->widthStep;
channelshsv = colimgbot->nChannels;
datahsv = (uchar *)colimgbot->imageData;
//————————–

heightmono = monoimgbot ->height;
widthmono = monoimgbot->width;
stepmono = monoimgbot->widthStep;
channelsmono = monoimgbot->nChannels;
datamono = (uchar *)monoimgbot->imageData;
cvCvtColor(frame,colimgbot,CV_RGB2HSV);
cvNamedWindow("original", CV_WINDOW_AUTOSIZE);

cvNamedWindow("Monochrome Of red Blob",CV_WINDOW_AUTOSIZE);
for(i=0;i< (heighthsv);i++)
{
for(j=0;j<(widthhsv);j++)
{
if((datahsv[(i)*stephsv+j*channelshsv]<=hlower) && (datahsv[(i)*stephsv+j*channelshsv]>=hupper))
{ if((datahsv[(i)*stephsv+j*(channelshsv)+1])>sthreshold){
datamono[i*stepmono+j*channelsmono]=255;}
else
/*A very simple concept with the loops here if the hue values are in the aforementioned range and the
threshold is met then logic one else logic zero*/
datamono[i*stepmono+j*channelsmono]=0;}
}}
for(i=0;i< (heighthsv);i++)
{
for(j=0;j<(widthhsv);j++){
if(!(datamono[i*stepmono+j*channelsmono]==0 || datamono[i*stepmono+j*channelsmono]==255))
datamono[i*stepmono+j*channelsmono]=0;}}/*Just a cross chek to ensure whether all the pixels have only
either 0 or 255*/
/*Please check these links for the explanation of the erosion and dilation functions

http://www.dca.fee.unicamp.br/dipcourse/html-dip/c9/s4/front-page.html*/

/*so now the last parameter in the function indicates how many times you want to apply dilation

or erosion*/

cvErode(monoimgbot,monoimgbot,0,6);
cvDilate( monoimgbot,monoimgbot,0,10);
/*here i have experimented with the values by changing them…and i have found
that i come to a good result by applying erosion 6 times and dilation 15 times
you can comment/uncomment play with the values and see what is going on
Sometimes you will find the areas which are shining in the image also get detected…

Please think why and then try to post a comment the best commment would get visible on this page*/

cvShowImage("original",frame);
cvSaveImage("red-ballmonochrome.jpg",monoimgbot);/*if you want to save the image*/
cvShowImage("Monochrome Of red Blob",monoimgbot);
cvWaitKey(0);
/*for all the other clarifications you can check the other posts…. you will find an answer
People who are getting started with the Opencv must make sure you

check the other posts on this blog*/

cvDestroyWindow("Monochrome Of red Blob");
cvDestroyWindow("original");
return 0;
}

3d-09-5balls