how to use matlab to smooth the edge of a picture

Question

I get a picture below but the picture has a very ugly edges. I want to use matlab programming to smooth the edges of the picture and make it look more pretty,any ideas or ways make sense?

thank you!!!

Answer 1

You could use imopen to morphologically open the RGB image (dilation and erosion). The second input argument to imopen function is a structuring element which defines the amount of smoothing required in the morphological operation. For example, below is a code where I apply a disk structuring element of radius 10.

img = imread('http://i.stack.imgur.com/ht4Za.jpg');
imopenBW = imopen(img, strel('disk',10));
imshow(imopenBW)

Answer 2

Here is an option (edited):

I = im2double(imread('ht4Za.jpg'));
% Segment the object:
gs = rgb2gray(I);
Object=~im2bw(gs, graythresh(gs));

% Smoothen the mask:
BW = bwmorph(bwconvhull(Object), 'erode', 5);
Mask=repmat(BW,[1,1,3]);

% Iterate opening operation:
Interp=I;
for k=1:5
    Interp=imopen(Interp, strel('disk',20));
end

% Keep original pixels, add the newly generated ones and smooth the output:
Interpolated(:,:,1)=medfilt2(imadd(I(:,:,1).*Object, Interp(:,:,1).*~(BW==Object)), [4 4]);
Interpolated(:,:,2)=medfilt2(imadd(I(:,:,2).*Object, Interp(:,:,2).*~(BW==Object)), [4 4]);
Interpolated(:,:,3)=medfilt2(imadd(I(:,:,3).*Object, Interp(:,:,3).*~(BW==Object)), [4 4]);

% Display the results:
Masked=imadd(Interpolated.*im2double(Mask), im2double(~Mask));
imshow(Masked);

Result:

It's a bit rough, but that'll give you a start. You can try to fiddle with the number of iteration and the size of the circular filter and the median filter. Try changing the median with average, etc.

Answer 3

You could use isocontour [1] edge detection and then average adjacent pixels along the contour to smoothen the boundary.

[1] http://www.mathworks.com/matlabcentral/fileexchange/30525-isocontour