How to remove the camera printed name by manufacturer in images using matlab?

Can anyone help me in getting to know the code for removing the camera printed name from manufacturer in any images?

[KSSV]

Attach one image.....and tell your expectations..

[ezhil K]

I just want the code for removing these unwanted words from image. In this image the words like "Shot on one plus" has to be removed. So I need matlab code for removing these unwanted words in images.

[Walter Roberson]

Please do not close questions that have an answer.

Answer 1

[KSSV]

This is one of the naive suggestion (may be). Off course Image Analyst's would be the best.

I = imread('1502388520041.jpg') ;

txt = ocr(I) ; 

R = txt.WordBoundingBoxes ;

imshow(I)
for i = 1:size(R,1)-1
    C = [9 118 183]/255 ;
    rectangle('position',R(i+1,:),'Facecolor',C,'EdgeColor','none')
end

YOu need to play around with R, to mkae it more perfect.

Answer 2

[Image Analyst]

The watermark will always be in exactly the same place. So you simply have to make a mask, like I did below:

and use it together with regionfill() to fill in the mask area with the surrounding color. Easy, but let us know if you have any problems you cannot solve.

[ezhil K]

I dont need these words.Instead I need to extract only the image by leaving these words.I couldnt understand the exact working of watermark.So could you please attach the matlab code for the same?

[Walter Roberson]

The mask that Image Analyst created shows the places that need to be replaced. You can create a mask from the camera_mask image file that Image Analyst provides, and then you can use that mask with regionfill() on one of your existing color images so that MATLAB fills those locations only.

[Image Analyst]

I'm sure you've looked up the regionfill() documentation to see the one-liner MATLAB code (maybe not though, since you're asking about it), but looks like you've accepted KSSV's answer and so it looks like you've found a solution that's good enough. Keep in mind though that it fills in the whole bounding box with a constant color so you are going to have edge effects with varying background, and it only works for that particular blue color. If the color is anything else, you'll have to determine a new C but you'll still have basically a uniform box on the image. But anyway, glad he was able to provide a solution that works for you. Though I would have done

newImage = regionfill(originalImage, mask);

instead. For a color image, you'll have to regionfill each color channel individually and recombine into the final RGB image.