Help eating cicchetti

I would really like to do a take on this PlayRaw (thanks @sguyader for the raw!). However, I am having trouble removing the vampire in the bottom left corner and doing perspective correction. I am not getting results that I like. Any advice would be appreciated.

1 Like

@afre regarding perspective correction, what do you want to achieve ? Getting vertical lines vertical?
Why do you want to remove the vampire, is it distracting to you?

I guess the point of this thread is to see the approach people would take and how successful they are with it. Despite my helpful posts in discuss, I am actually really bad a image editing!

Perspective When I compensate for one quadrant, the others look markedly worse. Maybe it is the wide angle and camera placement that is throwing me off.

The Vampire The scary face, white shirt, mostly solid body and vampiristic posture. Distracting indeed! Also, the couple doesn’t fit in my plans for the raw.

I took this picture on a bridge over the Canale, so the camera was looking down to get more of the interesting foreground. The field of view is not very wide, the 23mm lens on my APS-C camera is equivalent to that of 35mm on a “full frame” camera.
I’ll see what I can do for the vampire. Gonna be looking for Van Helsing I guess.

@afre, for the fun I’ll give away another shot of the same scene (released under Creative Commons Attribution, Non-Commercial, Share Alike license, aka CC-BY-SA-NC):
DSCF1720.RAF (20.9 MB)

No vampire in this one. Have fun!

A slight correction of perspective done in gimp. @afre, is that what you want?

Different from my tries. Still disorienting to look at, like I am seasick. What was your approach in GIMP?

Is it possible that the blurry/moving parts and the meaning poles make you disoriented?

I like the result of Camera Raw’s Transform Tool but I would rather use a FOSS solution.

Default settings + auto perspective correction

You should be aware of two things:

  1. The poles to which the boats are attached, are not perfectly vertical, and may play with our perception of perspective.
  2. The walls of the houses in Venice are not all perfectly vertical either.

@sguyader I am well aware but as I said earlier: when I correct the perspective, I like what it does to one part of the frame but not the rest. I guess PS can make multiple perspective adjustments. I tried dt because its tool looked fascinating but I am not familiar enough with it to get “good” results.

@Karl @paperdigits First one of 2018 - #5 by afre Please continue the hugin discussion here. I don’t want to take away from @matejmarti’s showcase. Say I would like to align Help eating cicchetti - #9 by afre and

(I want to straighten the second image with the first as reference.)

Could you give me some pointers on how to begin? I know what hugin does in principle but I get overwhelmed by it since I barely have any experience with it.

Before you begin, you should be aware that since the two images are taken from slightly different vantage points you’re going to get some parallax errors - alignment points in the foreground will lead to large errors in the background, or vice-versa, so you need to decide which is more important (place most of your control points in that area.)

Convert the images from RAW to TIFF using the same lens correction (if any) applied (ie. either have lens correction applied to both, or not at all.)

Start Hugin. If you’re using a version of Hugin newer than 2014, go to the “Interface” menu and make sure that either “Advanced” or “Expert” is selected, and that you’re in the window named “Panorama Stitcher”. The window should have for tabs (“Photos”, “Masks”, “Control Points”, and “Stitcher”.)

Click “Add Images” and load your photos. Right click on the image you want to align to and select “Anchor this image for position”.

Go to the control points tab, make sure that “auto fine-tune” and “auto-estimate” are set and create some control points (you don’t need many - 6 or 7 would be lots for the images here) - they should be only on static elements (sidewalk, buildings, etc.) and spaced around the image (there should be some in the corner areas, and near the center, if possible.) The best locations are areas that are perpendicular to the camera (these are less likely to suffer from parallax issues) but work with what you have (I used a single point on the sidewalk on the lower left, and a couple of the window corners and roof peaks.)

Go back to the “Photos” tab, and next to the “Geometric” pulldown, make sure “Positions (incremental, starting from anchor)” is selected, and hit the “Calculate” button next to it.

Hugin will calculate the position report how accurately it was able to do it - the numbers reported are in pixels. (Average should be less than 1 - if the maximum is much higher than the average + std. dev., then you probably have a bad control point.) Click “Yes” to apply the changes (and if you had any bad points, go back to the control points tab and delete and re-create them in a slightly different area, then come back and optimize again.)

Once you’re happy with the control points, go to the Stitcher tab and select “Rectilinear” projection, then hit “Calculate field of view”, “Calculate Optimal Size” and “Fit Crop to Images” . De-select every checkbox except “No exposure correction, low dynamic range”. If you want to preview your work go to the Preview tab of the “GL Fast Preview” window - you can see the overlap of the images, and toggle the images on and off to see what’s being done. (You can also visually adjust your crop if you want from the “Crop” tab.)

Go back to the “Stitcher” tab, hit “Stitch!”, and wait for Hugin to do its thing. When it’s finished you will have one remapped image for each source photo, named “huginprojectname_exposure_layers_XXXX.tif” - one will be a (probably cropped) version of the reference image, the others will be the other remapped images.

2 Likes

@Karl Each time I use Hugin, I learn that I have been using something wrong in the previous attempts; so thank you for your short guide. I only really use Enfuse CLI.

DSCF1719.jpg and s.jpg both come from the same raw file. My last and only successful attempt was using s.jpg by itself when I finally found out how the control points worked. However, it looked butt ugly probably because I need to understand Hugin more and apply your suggested settings.

Ahh, I thought you were looking to align DSCF1719 and DSCF1720 (posted in the original thread.) As you’ve probably guessed, you can disregard my preamble with regards to that. :slight_smile:

I hope you don’t get discouraged; Hugin is a pretty complex piece of software with lots of different uses - and it has a learning curve to match. :slight_smile:

@Karl I was too ambitious for a newbie, wanting to do 4 things in one go.

I also think that the GUI isn’t intuitive or easy to use; e.g., control points for horizontal and vertical lines. I had to search for various tools and settings; e.g., I struggled to find out where I could optimize and commit the adjustment. The optimize tab was greyed out but I later found that there was a button.

PS I haven’t made another attempt yet but I deleted the preferences and reinstalled. Turns out the leftover settings from previous versions were complicating matters.

1 Like