Page to fuse differently exposed images

[eengelmann]

I have now updated also the English version of my page which allows to fuse online two differently exposed pictures: http://www.milanofotografo.it/englishFotografiaHDR.aspx
In this way it is possible to get correctly exposed pictures of scenes which otherwise would result partially eighter too dark or burnt.
The page offers also the possibility to correct horizontal and vertical shifts, so to make it possibile to apply the algorythm also on pictures shot with the camera handheld.
There is also a help page.
Critiques and suggestions are welcome!
Note: If the images are larger than 12 Mp each, there could be an error!

The image below (detail inside the church of Sant Antonio Abate in Milan) was obtained using the algorythm contained in the page.

[IzzieK]

Thanks for the link...and welcome to this forum. Your webpage is very interesting. I will go back to it when my eyes goes back to normal. The dilating fluid my optometrist put in my eyes this afternoon is creating havoc to what I am looking at. I will be back to normal soon, I hope...

[eengelmann]

I'm glad you find my page interesting! In the future I will add more and more tools and options. But it takes time!
One thing I want to add is the possibility to calculate the average of several pictures, possibly subtracting from each the dark background, in order to be able to obtain good astronomic pictures even without following the star during the shot.

[eengelmann]

I have added some improvements to the page. In particular there is now a new gradient shape, expecially useful with images containing bright low contrast parts!

[Shadowman]

Thanks for the link.

[eengelmann]

New important update!
Now it is possible to start also from three differently exposed pictures, not only from two like before!
All the interested ones are invited to try it!

[eengelmann]

New update of the page: it is now possible to correct the shift also in case of three starting picture, not only in the case of two as before!
Try it out!
Critics and suggestions for future improvements are welcome!
The picture of the portico was obtained from three pictures on a scale of 2 stops taken with the camera handheld.

[chauncey]

Ah...why the 12MP limit and...what are the benefits over "merge to HDR"?

[eengelmann]

12 Mp: the problem depends from memory capacities of the server, not because of a limitation of the algorythm in itself. The pictures must be converted in arrays, which rapidly increase in length, with the pictures growing in size.
Since I only have cameras with sensors of 12Mp or less, for the moment I haven't addressed the problem.
benefits over "merge to HDR": What do you mean?

[eengelmann]

I have just added a tool which helps estimating the rotation differences between the bracketed pictures. There is also a dedicated instruction page: http://www.milanofotografo.it/englis...isiAngolo.aspx

[xpatUSA]

Ah...why the 12MP limit and...what are the benefits over "merge to HDR"?

What is "merge to HDR"?

[Inkanyezi]

Merge to HDR is an algorithm chosen to do exactly what is covered in this thread. It takes two or more images and merge them in a way that over-exposed (burned) highlights are disregarded as well as blocked shadows, if there are other images in the series that have rendition in those parts of the image.

After merging, the image will be too "deep" to show in the limited space of the computer, and if its tonal values are compressed into a more narrow space, it will become very soft. Therefore, before presenting the image, its global contrast should be retained, while local contrast, detail in the highlights and shadows as well as mid-tones, will be enhanced. The process for doing so is called tone-mapping. Typically a 32 bit image is compressed into a smaller bit-depth in that way.

The process can be done manually, but is easier with a well balanced algorithm.

In many cases, when the dynamic range of the camera is high, and there are no blocked shadows or burned highlights in the RAW image, tone mapping alone is sufficient to render the tonality well, although doing this process to one single image will lift noise and make it visible, while the multi-image method may render deep shadows without apparent noise.

Many HDR programs suggest using several images, but with modern cameras, you never need more than two.

[revi]

Could you tell us what is the advantage of your system over existing systems like e.g. the Hugin suite?
Any decent modern computer can run that suite with no problems for 20MP images at least (ok, there's a learning curve involved to get the best results, as usual)

[eengelmann]

I don't know if it is better, never used other tools.
One advantage could be that it is on line, so that you can use everywhere, if you have internet.
An other could be that the algorythm is described in detail (if the info on line is not enough I can give you more).
I have created mostly for myself, if someone likes it, it is there!

[DanK]

The process Urban described, High Dynamic Range processing, is the most common way to deal with too large a dynamic range, but it is not the only one. The other method that is fairly common now is "exposure fusion," which does no tone mapping. Exposure fusion simply combines well-exposed pixels from several images. One common software solution for exposure fusion is a Lightroom plugin called Lightroom Enfuse, which I use.

There are many HDR programs, and HDR capability is now built into both Lightroom and Photoshop. I don't have the right computer with me to check, but I believe "merge to HDR" is the lightroom command for invoking HDR.

[eengelmann]

Yes, I think my page works based on the same principles like exposure fusion!

[eengelmann]

I have just added to my page a new gradient, or, better, the possibility to let the system use for each pixel a mix of the two gradients, with the weights of them varying from pixel to pixel depending on the luminosity of the pixel itself and of the pixels surrounding it!
Furthermore there is now also a page which illustrates with some examples the different results you can achieve depending on the number of the starting images and the gradient shape chosen: http://www.milanofotografo.it/englis...HDRonLine.aspx

[eengelmann]

On line the new version of the tool for analysing the rotation of the starting pictures and correcting it and the corresponding instructions: http://www.milanofotografo.it/englishFotografiaHDR.aspx
This new version is faster and allows to handle more than just one border in the chosen area!

[inkista]

... One common software solution for exposure fusion is a Lightroom plugin called Lightroom Enfuse, which I use. ...

Footnotes:

The plugin is called Lr/Enfuse, and it's just a Lightroom GUI front-end for two line-command open source tools, align-image-stack (which lines up the member images), and enfuse (which does the actual fusing). There are actually a large number of GUIs that use these two tools to perform this task (e.g., Hugin).

Enfuse can weight how pixels are selected/combined from the member images by three factors: exposure, saturation, and contrast. If you weight to exposure, enfuse is exposure fusing. If you weight to contrast, enfuse is doing focus stacking. It's a pretty versatile tool, and the Lr/Enfuse plugin will let you adjust those weights however you like. But those sliders do not directly affect the exposure/saturation/contrast of the output image.

[eengelmann]

So focus stacking is based on contrast and just on that? Interesting! It means it is something I could do myself and add it to my page!