Teleconverter or Gigapixel?

JimH123 said:

I just read Phocal's response and I very much agree.

I am going to supply what has been asked for but hasn't been provided. First of all, I do not have a TC for my Olympus cameras. And probably never will.

But I do have several generic TC's in A-mount which I haven't used in a very long time, because I do not like the results I get from them. First one is a Tamron 1.4x. Second one is a Kenko 1.4x. Third one is a Kenko 2.0x. And finally, just to give one more example, I have an Olympus TCON-17x. This is a front end TC that screws onto the front of the lens, and is wonderful to use - that is if you don't mind all that extra weight hanging out there on the front of the lens.

For this test, I decided to use my Sony A6600 with the LA-EA5 adapter to allow it to work with A-Mount lenses. And I set the camera on a tripod, turned off Image stabilization and set the camera to f4. And I let the camera decide shutter speed and ISO. The lens was a Minolta 100mm f2.8 macro lens.

Considering this is late in the year and the days are short, and the fact I didn't start until 2:30PM, I had to use some tree branches high enough to still be in sunlight.

Here is a starting image with no TC and no Gigapixel AI. I loaded the images into Lightroom and used the default settings before exporting as a JPEG.


Original, no TC, No Gigapixel

Next, I am going to supply 3 images where I use Gigapixel AI (version 6.2.2) and I chose the LowRes Model. First will be 1/4x. then 1.7x, and finally 2.0x. The sliders for noise and remove blur were set very low, at 6 and 12 respectively as I didn't want to grow artifacts.

Note: I did not do any cropping, so the images will all look the same. But if you zoom in, there will be a growing number of pixels allowing it to be zoomed in farther before pixelization is reached.


Gigapixel at 1.4x

View attachment e7437c79be0141ab9efff7e7b70c793e.jpg
Gigapixel at 1.7x

View attachment 9440d42bc2c640949d74998a0757f641.jpg
Gigapixel at 2.0x

Now for the TC's. Now keep in mind that these are generic TC's, and they are not as good as dedicated TC's. But even dedicated TC's do degrade an image since you are spreading out the cone of light, and any lens problems are exasperated. I think that those of you who thought that TC's might be better than Gigapixel AI are going to be in for a surprise.

One more thing to point out. The EXIF information from the camera can have errors in reading the TC and may not show correct information for FL and aperture. The Tamron TC was worse than the Kenko TC's.


Tamron 1.4x


Kenko 1.4x


Kenko 2.0x

And here is the real surprise, the front-end Olympus TCON-17x. This can be explained in that the TCON-17x magnifies the image 1.7x before it goes into the lens. There is no spreading of the light cone. Just the tiny little problem of all that weight on the end of the lens. But no loss of light.


Olympus TCON-17x front-end TC

Perhaps I should also run that first image through ON1 Resize RAW 2025. I was playing with it last night and not finding it able to do as well as Gigapixel AI. If I get motivated, if someone wants to see the ON1 Resize results, I will add them.

I thought of pasting in side-by-sides, but the results between Gigapixel AI and the TC's are so startling different, that I didn't bother.

I will now go and put those TC's back into retirement.

*********************************************************

Update - here is something that the TC's can't provide: 4x and 6x

I did have to crop the image to do this as the images were growing too big to be supported.

View attachment 527c082e41bc41e2b77cd88d4bcd2f4e.jpg
Gigapixel AI - Cropped and resized 4x

View attachment a93662ee82cb41e08c6a481543812c96.jpg
Gigapixel AI - Cropped and resized 6x

Click to expand...

tomhongkong said:

J

My workflow at present is

import to DXO

Crop to intended size of final image

Basic curve/contrast adjustment

Turn sharpening off all together

Export in Deep Prime as JPG

Click to expand...

tomhongkong said:

Open with Gigapixel

Turn sharpening and other advanced settings off (I have more playing to do with this before I settle on the best setting)

Export as JPG after selecting desired size

Click to expand...

tomhongkong said:

Open with AI sharpening and process

Export

Final touch up with DXO

tom

Click to expand...

JimH123 said:

Based on your comment about comparing results on you small laptop, I decided to make it easier to compare results.

The image on the left is the original shot with no TC and increased to approx 200%. The image on the right is with Gigapixel AI and increased to match the same screen size.


No TC vs GPAI 1.4x


No TC vs GPAI 1.7x


No TC vs GPAI 2.0x

Next, the various TC options on the left and on the right, the corresponding Gigapixel AI image.


Tamron 1.4x vs Gigapixel AI 1.4x


Kenko 1.4x vs Gigapixel 1.4x


Kenko 2x vs Gigapixel 2x


TCON-17x vs Gigapixel 1.7x

Let me know if you want to see any other comparisons.

What is unknown at this point is how much improvement there would be if TCs designed for a specific lens were to be used instead of these generic TCs which as we can see, just aren't very good. In fact, the Olympus TCON-17x front end TC performed better than these generic TC's, but not as good as Gigapixel AI.

************************************************

Adding one more comparison. Gigapixel AI at 2x on the left vs ON1 Resize RAW 2023 at 2x on the right


Gigapixel 2x vs ON1 Resize 2x


No TC vs GPAI 4x


No TC vs GPAI 6x

My final thought on this is that Gigapixel AI really works - once you figure out how to use it without artifacts and halos. But if you want to see artifacts and halos, I do know how to create them!

My advice:

- Start with the best image you can create with your camera. Sharp focus, and no camera shake.

- And when using GPAI, be easy on the sliders. Keep them to the left and ever so gently, move them to the right only until you see artifacts and halos at high magnification, and immediately back off until they are gone, and go a little farther to the left just to make sure. GPAI does remove some noise, but it is better to remove the noise before you even get to GPAI.

Notice the order they are done in PhotoAI -- first the noise, then the sharpening and finally the resizing.

Click to expand...

tomhongkong said:

I did try a TCON 17 on my Stylus 1 and didn't really get results which justified killing its size advantage (it makes it very cumbersome) I think I tried it on 75-300 as well but didn't get good results. I have since sold it. I have a few other front converters which I will try when I have time.

Cheers

tom

Click to expand...

tomhongkong said:

Thanks, that's very helpful. Your tests really show that GP easily beats the TCs, in your examples, but we need good results with better M4/3 lenses and TCs. My 'new' panasonic 200 f2.8 and 1.4x TC arrived today in a rainstorm in London. I could not get anything remotely decent in the gloom. Hopefully the weather will be better over the next few days and I will try 200mm, 200+ 1.4 TC, 200+1.4TC +1.4 resizing and 200 +2.0 resizing. That should give us better pointers.

Click to expand...

JimH123 said:

tomhongkong said:

I did try a TCON 17 on my Stylus 1 and didn't really get results which justified killing its size advantage (it makes it very cumbersome) I think I tried it on 75-300 as well but didn't get good results. I have since sold it. I have a few other front converters which I will try when I have time.

Cheers

tom

Click to expand...

I also have the Minolta ACT100 which is a 1.5x frontend TC. I find it less sharp than the TCON-17x.

These frontend TC's have several strikes against them:

1. They are big and heavy

2. The camera isn't aware of them, and the image stabilization setting has the wrong focal length.

3. When used on a lens with a larger diameter objective, step down rings may vignette the image. I find all my longer lenses have large diameter objectives.

*********************************************

I do have the Olympus 75-300mm, but haven't tried it with the TCON-17x.

But here is a test I did using a fourthirds 70-300mm lens with the TCON-17x. I thought I was at 300mm, but the weight must have caused it to retreat to 226mm. The second image used DenoiseAI and then GPAI at 4x. The gazebo is about 1 mile away (1.6Km).

I no longer have this lens as I decided I like the native Olympus m4/3 lenses so much better than using an adapter.

I also tried comparing this to the fourthirds 1.4x TC and the results were worse than using the TCON-17x. I sold that TC too.


226mm with TCON-17x


same image treated with Denoise AI then GPAI 4x

Click to expand...

OutsideTheMatrix said:

JimH123 said:

tomhongkong said:

I did try a TCON 17 on my Stylus 1 and didn't really get results which justified killing its size advantage (it makes it very cumbersome) I think I tried it on 75-300 as well but didn't get good results. I have since sold it. I have a few other front converters which I will try when I have time.

Cheers

tom

Click to expand...

I also have the Minolta ACT100 which is a 1.5x frontend TC. I find it less sharp than the TCON-17x.

These frontend TC's have several strikes against them:

1. They are big and heavy

2. The camera isn't aware of them, and the image stabilization setting has the wrong focal length.

3. When used on a lens with a larger diameter objective, step down rings may vignette the image. I find all my longer lenses have large diameter objectives.

*********************************************

I do have the Olympus 75-300mm, but haven't tried it with the TCON-17x.

But here is a test I did using a fourthirds 70-300mm lens with the TCON-17x. I thought I was at 300mm, but the weight must have caused it to retreat to 226mm. The second image used DenoiseAI and then GPAI at 4x. The gazebo is about 1 mile away (1.6Km).

I no longer have this lens as I decided I like the native Olympus m4/3 lenses so much better than using an adapter.

I also tried comparing this to the fourthirds 1.4x TC and the results were worse than using the TCON-17x. I sold that TC too.


226mm with TCON-17x


same image treated with Denoise AI then GPAI 4x

Click to expand...

Jim please try the TCON-17 on the 75-300 lens at 300mm and set the fnumber to f/9 as that is max resolution on that lens. Can you test that vs without any TC vs x2 DTC in camera and also the combo of the TC and the DTC together? From the reports I looked at, the TCON only helps for close subjects.

Click to expand...

OutsideTheMatrix said:

Jim please try the TCON-17 on the 75-300 lens at 300mm and set the fnumber to f/9 as that is max resolution on that lens. Can you test that vs without any TC vs x2 DTC in camera and also the combo of the TC and the DTC together? From the reports I looked at, the TCON only helps for close subjects.

Click to expand...

tomhongkong said:

OK, I am not necessarily advocating Gigapixel but it is pretty popular upsizing software. There are others.

However this is an interesting question. Assuming you had a 300f4 for example, and the bird image from 300mm is too small (low res) after cropping, are you better off using a TC or to upsize? Let's assume a 1.4xTC, the most commonly used size. (Bigger magnifications run into problems with both approaches)

The TC loses 1 stop of light. Gigapixel introduces artifacts as a result of upsizing (but not much with just a 1.4x upsize) Often I am balancing iso and shutter speed and 1 stop makes quite a difference.

Is 1 aperture stop loss worth the image degradation from upsizing.?... that's iso 1600 to 3200, or 3200 to 6400. DXO is getting pretty near the limit! How much does the 1.4x TC also introduce degradation? perceived wisdom is that the 2x TC does do so quite a lot.

What is your experience? Any samples?

Full disclosure: I am trying to decide if an MC14 is really an advantage on my 300, (or even an MC20!) or if I am better off doing it with software

tom

Click to expand...

tomhongkong said:

Phocal

Thank you for taking the time to post all of these. I know how much effort it takes to do a decent job of camparison and posting and appreciate your efforts. At last, someone is providing evidence to allow me (and others) to make the comparison. I am trying to do just that now but on my small laptop, I must say that at first glance there is surprisingly little difference between any of the samples, except that the upsized ones retain the original pixel count and I assume would be better if cropped further or printed large.

I will spend more time later today and comment again.

Once again many thanks

tom

Click to expand...

Phocal said:

tomhongkong said:

Phocal

Thank you for taking the time to post all of these. I know how much effort it takes to do a decent job of camparison and posting and appreciate your efforts. At last, someone is providing evidence to allow me (and others) to make the comparison. I am trying to do just that now but on my small laptop, I must say that at first glance there is surprisingly little difference between any of the samples, except that the upsized ones retain the original pixel count and I assume would be better if cropped further or printed large.

I will spend more time later today and comment again.

Once again many thanks

tom

Click to expand...

You never state what your purpose is for the photographs. I assume, since you mention upsizing, that it is for printing.

If it is just for posting to social media, it really doesn't matter. I say that, but if you are trying to crop away 3/4 of the photo to make a bird the size of a fly in the frame fill the frame... in that case, the TC will be better. But in more cases, you are better off just upsizing, and honestly, Gigapixel is just overkill for that.

Printing is a bit different. If you are only looking to print around 16x12, upsizing will produce a print that, for the most part, will be just as good as an uncropped image. But if you want to print large, and by large, I mean 30x20 or 40x30, I personally think there will be a quality of print difference.

I have a 40x30 on my wall that I took with the 300/4 and MC-14 that looks amazing. You can walk up to it and see the individual strands of feathers. This photo has zero crop on it. I have a very high standard for my prints. I know there are recommended viewing distances based on the size of a print. But, if you have a huge print on your wall, your friends will walk right up to it to inspect the image's details. It's like the whole internet thing where people can zoom in on a photo makes them walk up to a huge print and expect the same detail they see when zooming in on the computer. So, I want my large prints to have that kind of detail. I personally don't think if I had taken the image with my bare 300/4, the detail wouldn't hold up.

This is why I always try to get the photo right in the camera. Which, for me, includes not cropping. Probably 95% of the images you see me post have less than a 10% crop on them. I came from full frame and had pixels to throw away and still be able to print large. When I moved to Olympus, I knew things would change with respect to large prints. I don't have the pixels to throw away and now have to get it as perfect in camera as I can.

If you want to use the argument about how great these new programs are (personally, I feel they are not as great as everyone says) and you just enlarge it for bigger prints. I will just counter with, just think, if you filled the frame with the subject, how much bigger could you print using the same software?

So, if the final use is the internet or small prints, just crop and upsize. If it is large prints, get the TC.

my two copper pieces

Phocal

Click to expand...

tomhongkong said:

Phocal said:

tomhongkong said:

Phocal

Thank you for taking the time to post all of these. I know how much effort it takes to do a decent job of camparison and posting and appreciate your efforts. At last, someone is providing evidence to allow me (and others) to make the comparison. I am trying to do just that now but on my small laptop, I must say that at first glance there is surprisingly little difference between any of the samples, except that the upsized ones retain the original pixel count and I assume would be better if cropped further or printed large.

I will spend more time later today and comment again.

Once again many thanks

tom

Click to expand...

You never state what your purpose is for the photographs. I assume, since you mention upsizing, that it is for printing.

If it is just for posting to social media, it really doesn't matter. I say that, but if you are trying to crop away 3/4 of the photo to make a bird the size of a fly in the frame fill the frame... in that case, the TC will be better. But in more cases, you are better off just upsizing, and honestly, Gigapixel is just overkill for that.

Printing is a bit different. If you are only looking to print around 16x12, upsizing will produce a print that, for the most part, will be just as good as an uncropped image. But if you want to print large, and by large, I mean 30x20 or 40x30, I personally think there will be a quality of print difference.

I have a 40x30 on my wall that I took with the 300/4 and MC-14 that looks amazing. You can walk up to it and see the individual strands of feathers. This photo has zero crop on it. I have a very high standard for my prints. I know there are recommended viewing distances based on the size of a print. But, if you have a huge print on your wall, your friends will walk right up to it to inspect the image's details. It's like the whole internet thing where people can zoom in on a photo makes them walk up to a huge print and expect the same detail they see when zooming in on the computer. So, I want my large prints to have that kind of detail. I personally don't think if I had taken the image with my bare 300/4, the detail wouldn't hold up.

This is why I always try to get the photo right in the camera. Which, for me, includes not cropping. Probably 95% of the images you see me post have less than a 10% crop on them. I came from full frame and had pixels to throw away and still be able to print large. When I moved to Olympus, I knew things would change with respect to large prints. I don't have the pixels to throw away and now have to get it as perfect in camera as I can.

If you want to use the argument about how great these new programs are (personally, I feel they are not as great as everyone says) and you just enlarge it for bigger prints. I will just counter with, just think, if you filled the frame with the subject, how much bigger could you print using the same software?

So, if the final use is the internet or small prints, just crop and upsize. If it is large prints, get the TC.

my two copper pieces

Phocal

Click to expand...

Thanks for your continued interest in my question. In answer to yours, my objective is to get better prints of distant wildlife, I used to say that I was most interested in A4 size (after cropping) which are easy to store, but now I am thinking larger, and will probably get an A3+ printer when I get home to print wall images (the selection of which I will have to rotate as I don't have much wall space)

The drivers for my question are

1. How to get longer reach without increasing size/ weight (PL 100-400 is already stretching my ability to hike, at the age of 83) Getting closer to birds is frequently just not possible for me, much as I would like to get a larger image in the frame.

Click to expand...

tomhongkong said:

2. How to keep the expenditure on lenses within reason

3. The claimed improvements in AI software based image enhancements.

You and one other poster have been kind enough to respond to my post as I hoped by giving examples of both TC and software 'magnifications'. Based on these postings, I agree with your comment above that for printing, at least in moderate sizes up to A3+ Gigapixel looks as though it will do the job perfectly well.

I have been doing further testing myself with Gigapixel. It really seems to me to be very unpredictable in terms of output quality and artifacts. If the image is busy, artifacts are easily lost and it looks good. If the background is clear, artifacts can be seen very easily particularly if the image has high cotrast straight lines. My own test subjects have been aerials on rooftops and the cables to them, and they show artifacts almost immediately from the lowest setting in Topaz. It is still a work in progress for me to find out the best Topaz settings.

(I have been using PL200 with or without 1.4TC for my tests, and comparing with PL100-400)

Thanks for your ongoing help

Click to expand...

JimH123 said:

tomhongkong said:

J

My workflow at present is

import to DXO

Crop to intended size of final image

Basic curve/contrast adjustment

Turn sharpening off all together

Export in Deep Prime as JPG

Click to expand...

Why export as JPG here. Gigapixel works much better with DNG or TIF files.

JimH123 said:

Open with Gigapixel

Turn sharpening and other advanced settings off (I have more playing to do with this before I settle on the best setting)

Export as JPG after selecting desired size

Click to expand...

Same comment here. Why are you using JPG at this step?

JimH123 said:

Open with AI sharpening and process

Export

Final touch up with DXO

tom

Click to expand...

Make conversion to JPEG the very last step.

Click to expand...

tomhongkong said:

That didn't seem to be the clearest way of showing the relative performances. I am trying again, with the actual size of the images (actually all cropped to equivalent of 2000x1500 to make it a bit more obvious)

I hope that makes more sense, otherwise please just ignore! Incidentally all settings on Gigapixel are at the lowest. I just find it does not do well where there are strong contrasts ad sharp edges like the lettering in these samples. It looks OK to me on the bricks

Based on these tests, I will continue to use the 1.4TC, and be very cautious of using GP to get a bigger magnification if I am hoping to print really large. I guess I will have to try to find a 2.0TC! (or stick with my 100-400)

Note, these were taken in atrocious conditions in RAW processed in DXO without any sharpening

toRAWI

Click to expand...

JimH123 said:

My final thought on this is that Gigapixel AI really works - once you figure out how to use it without artifacts and halos. But if you want to see artifacts and halos, I do know how to create them!

My advice:

- Start with the best image you can create with your camera. Sharp focus, and no camera shake.

- And when using GPAI, be easy on the sliders. Keep them to the left and ever so gently, move them to the right only until you see artifacts and halos at high magnification, and immediately back off until they are gone, and go a little farther to the left just to make sure. GPAI does remove some noise, but it is better to remove the noise before you even get to GPAI.

Notice the order they are done in PhotoAI -- first the noise, then the sharpening and finally the resizing.

Click to expand...