How to deblur an image where the text is blurred by imgaussfilt() function?

Currently, we have built our ID text blurring code using ocr() to detect all the text found on the image file then create a gaussian blur mask for each line of text using poly2mask from the values stored in bboxes.

Here's the code:

%Read Image

ID = imread('passport1.jpg');

imageSize = size(ID);

h = fspecial('disk', 3);

%Detect all text on ID

ocrResults = ocr(ID);

bboxes = ocrResults.WordBoundingBoxes(:,:);

[x,y] = size(bboxes);

for i = 1:x

xi = horzcat(bboxes(i,1), bboxes(i,1), ...

bboxes(i,1) + bboxes(i,3), bboxes(i,1) + bboxes(i,3));

yi = horzcat(bboxes(i,2), bboxes(i,2) + bboxes(i,4), ...

bboxes(i,2) + bboxes( i,4 ), bboxes(i,2));

%Convert Region of Interest to region mask

binaryImage = poly2mask(xi, yi, imageSize(1), imageSize(2));

%Use gaussian filter w/ 10 standard dev for larger filter

blurredImage = imgaussfilt(ID,20);

%Mask with 'same' option to retain size of

%output array from input array

blurredMask = imfilter(im2double(binaryImage),h,'same');

%Rescale image

ID = im2uint8(im2double(blurredImage).*blurredMask + im2double(ID).*(1-blurredMask));

end

imshow(ID);

We tried functions such as deconvlucy():

luc1 = deconvlucy(ID,h,15);

imshow(luc1);

and deconvreg():

noiseMean = 0;

noiseVar = 0.01;

blurred_noisy = imnoise(ID,'gaussian',noiseMean,noiseVar);

noisePower = noiseVar*numel(ID);

Edged = edgetaper(blurred_noisy,h);

reg3 = deconvreg(Edged,h,noisePower/1.2);

%subplot(133); imshow(reg3);

figure(2); imshow(reg3);

---which are added after the for loop so we can immediately see how the image was deblurred.

Here are the images:

Start

Blurred Text

deconvlucy()

and deconvreg()

After using the deblurring functions, it looks like it did little to no work on the filters

I would like to ask if there's a way to reverse the imgaussfilt() function that was used for the text blurring

Attached is the starting image file as well.

Thank you in advance!

[Martin Adrian Enghoy]

@Image Analyst We would like to ask for your help on this one! :D

Answer 1

[Image Analyst]

There's no way. The information has been totally lost and obliterated. No amount of deblurring will restore that in an image that has been converted to uint8 and saved to an image file. Perhaps if you had the original blurred image in floating point, but even there, I doubt it. This is real world, not Hollywood:

No Way Out https://www.youtube.com/watch?v=m4SBMyd0yEQ

(Most hilarious misuse of image processing in movies)

Answer 2

[ali hassan]

what is the benifit of bluring and what are its application?

[Image Analyst]

Well, it can obscure details in a picture, as you can see from your example. You can also use it to estimate a background in an image that is mostly background but with small foreground parts.