Background Removal Via Image Segmentation\ Threshold technique

In this post, I would like to discuss about how to remove background using image Segmentation/thresholding technique.

Image segmentation or thresholding  is the process to separate the foreground region and the background region in the image. The foreground region is the area of interest which refers to the area which contains the object. The background region refers to the region which is outside of the interested areas. The background region does not contain any important or valid information. For image segmentation we used color thresholding in l*a*b color space (depends upon you whatever color you want to use). First we covert RGB color space to l*a*b color space then we defined the minimum and maximum threshold value for each channel in l*a*b color space. Then we have created a mask based on threshold values and convert image to black and white format. The white pixels in black and white image are restored to RGB color format. Now the output image will consist of object(s) and the background is removed. The Figure 1(a) shows input image and figure 1(b) shows black and white image that we got after performing image segmentation, in this image the white region indicate area of interests and white region indicate uninterested area that need to be removed. Figure 1(c) shows that the resultant image with background removed which will help to find the objects in image very easily.

Description:

Fist we create a separate MATLAB file main.m. Inside main.m we read the input image and display it as shown in the figure 1(a).

I=imread('download.jpg');

imshow(I);

Then we call the backround_removal.m function and pass the input image as parameter to the given function. The backround_removal.m function then return two images bw and rgb, the bw is a binary image in which the white region shows area of interest and the black region shows uninterested area which we want to remove as shown in the figure 1(b). The rgb is a color image which contains only the foreground region as shown in the figure 1(c).

bw, rgb]=backround_removal(I);

figure, imshow(bw);

figure, imshow(rgb);

Then we create another file backround_removal.m. Inside backround_removal.m we create the function named creat_mask which will return two output which are binary images and the masked RGB image.

function [BW,maskedRGBImage] = createMask(RGB)

Inside this function  we convert RGB image to chosen color space.

RGB = im2double(RGB);

cform = makecform('srgb2lab', 'AdaptedWhitePoint', whitepoint('D65'));

I = applycform(RGB,cform);

Next Define thresholds for channel 1 based on histogram settings

channel1Min = 0.392;

channel1Max = 99.927;

Similarly we define thresholds for channel 2 based on histogram settings

channel2Min = 18.196;

channel2Max = 77.530;

Similarly define thresholds for channel 3 based on histogram settings

channel3Min = -35.069;

channel3Max = 68.943;

Then we create mask based on chosen histogram thresholds

BW = (I(:,:,1) >= channel1Min ) & (I(:,:,1) <= channel1Max) & ...

    (I(:,:,2) >= channel2Min ) & (I(:,:,2) <= channel2Max) & ...

    (I(:,:,3) >= channel3Min ) & (I(:,:,3) <= channel3Max);

Then we initialize output masked image based on input image.

maskedRGBImage = RGB;

Then we set background pixels where BW is false to zero.

maskedRGBImage(repmat(~BW,[1 1 3])) = 0;

Results:

Figure 1(a): Input Image

Figure 1(b): Background Removed Image

Figure 1(c): Binary Image with Background Removed

Full MATLAB Code:

1. main.m

I=imread('download.jpg');

imshow(I);

[bw, rgb]=backround_removal(I);

figure, imshow(bw);

figure, imshow(rgb);

2. background_removal.m

function [BW,maskedRGBImage] = createMask(RGB)

% Convert RGB image to chosen color space

RGB = im2double(RGB);

cform = makecform('srgb2lab', 'AdaptedWhitePoint', whitepoint('D65'));

I = applycform(RGB,cform);

% Define thresholds for channel 1 based on histogram settings

channel1Min = 0.392;

channel1Max = 99.927;

% Define thresholds for channel 2 based on histogram settings

channel2Min = 18.196;

channel2Max = 77.530;

% Define thresholds for channel 3 based on histogram settings

channel3Min = -35.069;

channel3Max = 68.943;

% Create mask based on chosen histogram thresholds

BW = (I(:,:,1) >= channel1Min ) & (I(:,:,1) <= channel1Max) & ...

    (I(:,:,2) >= channel2Min ) & (I(:,:,2) <= channel2Max) & ...

    (I(:,:,3) >= channel3Min ) & (I(:,:,3) <= channel3Max);

% Initialize output masked image based on input image.

maskedRGBImage = RGB;

% Set background pixels where BW is false to zero.

maskedRGBImage(repmat(~BW,[1 1 3])) = 0;