It is hard to answer without looking at an image and Getting exactly the same image is not possible in most cases but you can use Morphological dilation to thicken your image. 

PS: Low pass filtering may also help in some cases. 

The image thinning function in MATLAB uses the values of the eight neighbours of a point “p”, and the process iterates until the image stops changing. The shape of the original object is, in most cases, not preserved but the Euler number is. The shape can be reconstructed with a single or iterative morphological dilation if it is a simple structure, say a straight line. Else, the operation is irreversible since the used information from the eight neighbours is lost. Note also that image dilation yields a shape that depends on the nature of structuring element being used.

Spatially or density?  In Photoshop density can be built by duplicating the layer and put the top layer in Multiply Blend Mode.  Spatially (image dependent) Select and Stroke area of interest.  Hope that helps!  Jim

Jim,

I am not so sure if your answer addresses the problem in question. Thinning is a morphological operation which removes selected foreground pixels from binary images and produces another binary output, therefore,  I failed to see the link to density.  I have attached an example image of thinning (hopefully it will show up).

I see, in that case we have a different issue than what I envisioned.

Thank you all for your response,

Abbas Cheddad, you are right about thinning. That it is irreversible, but still there must be a way to restore it.

And you said some thing about Euler number. Can't we use this euler number to thick image?

• Yes Jim, the term seems to have different meaning in your scientific field.
• I am afraid not Muhammad. The Euler number in binary image processing depicts the number of holes to the total number of objects present in a given image. I don’t see how that can be helpful in your case.

One can always attempt to deduce quasi-reversible implementations of irreversible transforms but that would work only, as I said above, in simple structured objects. The attempt for sure will not produce the desired output for complex topologies. Take for example the colour-to-grayscale image transform (3D to 2D), it is irreversible though they are some machine learning based algorithms assisted with lookup tables which create some interesting colour blend on a grayscaled image but that will not reconstruct the original vibrant natural colour of the image. This is because the rich information that once existed in the RGB colour space is lost and so is the case in your scenario.

Thank You for your explanation.

You are welcome Jim!

Thinning is indeed not reversible.

Two comparable algorithms are 'skeletonisation' and the 'medial axis transform'. Both are well defined operations in the continuous domain, but much less so in the discrete domain. If the output of such an operation, is not only the medial axis, but also a value (the distance to the original edge) for each point of the medial axis, then you can  recuperate the original image. For each medial axis point, generate a filled circle with a radius of its distance to the original contour. As said, there will be some losses because the medial axis transform is ill defined in the discrete domain.

The topology of the input blob is unimportant for the medial axis transform; the topology will be preserved.

Means same goes for skeletonisation. it is irreversible.

Thank you all for your help.

No, that does not mean that at all.

The medial axis transform is reversible. That being said, most implementations of a medial axis, actually make an approximation of that axis. If you use that axis to recuperate the original shape, you only get an approximation of that shape.

Take a filled circle for example. The medial axis is the center point with an associated radius. You can exactly recover the original shape. However, if the circle is digitized, you see pixelation of the rim. The exact medial axis is not the center point anymore, but would more look like a cob web extending to all corners of the pixelated rim!