Canon adds Sensor Shift Hi-Res to R5

Canon’s latest firmware update for the EOS R5 included a new feature they call IBIS High Resolution. This is similar to Sony’s ‘pixel shift’ hi resolution system introduced a few years ago in the a7R IV. Sony takes a 61 mp sensor and yields 4 or 5 images that are stitched together to form a single 250mp image, the images are all slightly offset using the IBIS system to produce a very hi resolution image. Sony’s system works pretty well with the caveat that you camera is firmly mounted to a sturdy tripod and the subject is still. Canon has one-upped them by giving a 400mp image after taking nine 45mp stills and generating one giant jpeg file. Have they really one-upped them? Read on.

The Canon system has to take roughly twice as many images to yield that larger file size and in that time period there is a greater chance of subject movement which causes softness and or artifacting, neither of which we want. Subject movement can be as benign as a slight breeze moving tree leaves. At 400 megapixels the final image will show ridiculously minute details. Canon did not enable this feature to use RAW files which is unfortunate but I think I understand why; 9 RAW files stitched together would be slower and the output file would be ginormous. Still, I’d like to option to go RAW.

My experience using the feature called IBIS High-resolution shot in the menu is mixed. I did a shot in my home office/man-cave of some cameras up on the shelf that yielded a brilliantly detailed shot. I was amazed and immediately started planning my downtown panorama shot using the feature. My results outdoors along the breezy south shore of the Columbia River looking north towards Downtown Vancouver, WA were not very good. I shot the images through my Linhof Technika III Mk V with a Schneider Symmar 135mm using my Multi-image adapter. I took three of the 400mp images and stitched them together in Photoshop to produced a single panoramic gigapixel image. I also took a few with the R5 mounted with the RF 14-35 L stopped down to f/8. None of them looked as crisp as the shot in my office. It wasn’t that the trees were blurry, that would be fine considering the breezy conditions. The high rise buildings were soft as well. Perhaps the winds buffeting the camera was enough to cause an issue, but honestly it wasn’t that windy, maybe a 10mph breeze with 15mph gusts. Hardly blowing a gale.

I took an image with the 14-35 L at f/8 and 29mm using a single 400mp IBIS hi-res shot. I compared it to a shot from the same location a year prior using 12 stitched 45mp images and a 135 Symmar. Both images are similar magnification with the Symmar being slighter more magnified but the final cropped panoramic are both about 150mp mp. The stitched 45mp image blow the doors off the 400 mp single shot. It is not even close. There is a some weird artifacting going on along with less detail. This hi-res option is not a replacement for a stitched panoramic, and honestly I did not expect it to be. I did however expect better results than I got.

A single shot 400mp IBIS hi-res cropped down to about 150mp to create skinny pano.

1:1 actual size IBIS hi-res

Stitched 45mp images 1:1 actual size

Here is the 1:1 blow up. You can see that it is hard to read the logo on the Indigo Hotel and the shot is just not that sharp. There is a weird artifact on a magazine box near the bottom left of the 1:1 tight cropped shot.

Comparatively the stitched 45mp shot done the “old way” offers significantly more detail. These were shot a year apart so the two buildings on the right were still under construction for the first image and are completed by the second image. To be fair, both images are edited using my standard presets for color and contrast. The contrast difference was due to different weather conditions on the two shooting days.

There is an advantage to using the longer lens and stitching images together and that is clear in the comparison of the two 1:1 crops. But I tried doing a three image set of hi-res 400mp stitched with each image using the same Linhof with the Symmar 135 and the image is soft well before zooming in to 1:1. It already starts softening up at 1:4 so the advantage of the extra megapixels is nil.

I took several other 400 mp images in other parts of the city and found all of them to be lacking detail at 1:1 resolution. All of them while mounted to my large tripod and with shutter speeds above 200th of the second. I think Canon needs to tweak this system bit. Hopefully they added this to the R5 to act as a beta field test before launching the feature on the next generation camera(s) likely the R5 Mk II in a couple years. I hope they give me a firmware tweak as well 🙂

For the record I did achieve very good results in my house of a still life. At 1:1 it wasn’t super sharp but much better than the images I made outside. I believe the concept is sound but Canon may have bit off a little more than it could chew going from 45mp to 400mp where as rival Sony does a more modest 61mp to 250mp. The Sony pixel shift system from what I have seen definitely performs better than the Canon IBIS hi-res. I’ll be anxiously awaiting a firmware update to make this feature more usable.

Don’t forget the next show is June 24th at Seawood Photo. Admission is free for this outdoor event in San Rafael, CA. As always check the website http://www.photofair.com for more information.

Read Full Post »

Composite Image Landscapes

How’s that for a title? Are you scratching your head? What the hell is he talking about? Well a composite image is simply using two or more images combined to make one image. This procedure is used in a wide variety of shots from simple sky replacement to dynamic astro shots. The process has allowed us to make amazing images using the sophisticated software that is all around us and today’s modern imaging sensors.

I do a lot of panorama work using up to 36 images to produce a wide panoramic image. This is a fairly simple process these days, in fact Photoshop does a good job of compositing panorama images without the need for expensive plugins. I even use some of my vintage gear to do this. I have a Linhof Technika III v5 in 4×5 with a Schneider Symmar 135/5.6. I use a special multi-frame camera adapter to mount my R5 to the camera and then make up to six images moving the assembly around to cover what is the equivalent of 44×88 cm. I wrote about that unit in a post, here. Photoshop then takes the images and puts them together blending them into one big panoramic shot. You end up with amazing details that just can’t be matched with a single shot. To get the image size in one shot I’d need a 50mm lens. With this panoramic shot I’m using a 135mm lens that has excellent resolving power. The details are amazing. The full resolution of the image below is 39743×5298 pixels or about 210 megapixels due to space limitations on this server it is show at far less resolution. You can barely see the red square. The cropped red square part blown up is 100% however.

Sometimes I just take a series of vertical shots with a lens at about 70mm and stitch those together. That requires rotating the camera which induces some distortion, but can be corrected in Photoshop. This is a less complicated and quicker process since using a 4×5 field camera is a bit slower than using the R5 with native lenses. I shoot vertical so I get more height to the image which allows for using a longer lens and thus getting more details. To composite them together, Photoshop now offers a tool built in to the software call “photomerge,” this is found under ‘file’ using automate which opens an option for photomerge. That opens a window that lets you choose a group of images to merge together. You can choose to force photoshop into a particular projection style but generally selecting “auto” here works well. I let the software correct vignetting. If you did it right, you won’t have much anyway.

You may see astro images where there is this amazing sky showing off the nebulous arms of the Milky Way with a spectacular foreground image of mountains, or rocks and such. These are often composites as well with the foreground being dropped in front of the sky. This is done by masking the foreground and creating a transparent sky then dropping image in front of the astro-sky shot to create the composite image.

Even shots of skylines with a dramatic sky can be made this way by shooting the sky separate from the foreground image best mountains, oceans, cityscapes or whatever. I have used this technique numerous times in my website header images. The latest versions of Photoshop have some free art skies you can even use if you don’t have your own dramatic sky shot. In some of my real estate marketing materials I have use a Vancouver city skyline shot that is a composite because I am unable to get my camera 100 feet off the ground on Hayden Island across from Downtown. So I take the city skyline shot from the opposite bank of the Columbia River, then I go to North Portland at roughly the same angle, I shoot Mount Saint Helens with a long tele lens from about 100-150 feet in elevation in Forest Park. In the original shot from the river bank only the very top of the mountain is visible in the scene due to my low elevation. Foreground hills cover up the lower slopes. From the higher position I can see the lower slopes peeking out over the local hills. It is a more dramatic view and is the view you get when using a drone. I mask out the sky and the mountain from the skyline shot and add in the mountain from the higher perspective to end up with a realistic looking representation as if I shot the picture from across the river in a high rise building that doesn’t exist. Clever tricks that save me the cost of renting a helicopter to get the shot. In the shot above which I use on my Facebook page I left the original sky in place but added a semi-transparent color filter to give it a dusk feel. Mount Saint Helens appears a little more prominent in the background as if I took the shot from 100 feet off the ground.

So today I want to discuss how to do this. If you are using the multi-frame 4×5 adapter that sets the shots up for you with an appropriate levels of overlap, take a look at that post for more info. Today I’m focusing on stitching and blending images made in your camera.

The general rule for multi frame panoramas I find to be the best, is to come up with one exposure for all the images. Do not do this on a day where the sun is ducking in and out of the clouds. As you move your camera to take the shots your light meter will get slightly different exposure readings that will affect the sky and make blending difficult. Take the camera and pan it along the intended photo path and look at the exposure readings. Then choose a happy medium, set the camera to manual and use that exposure for all the shots. This will make the sky very uniform across the images and minimize blending issues.

The next step is overlap. You should have at least 20% overlap of the images to allow the software to find adequate stitching points. This will limit the errors in stitching. A tripod is highly recommended to keep the images aligned on the vertical plane so as to avoid excessive cropping and additional stitching errors. I have made successful panoramas with out the tripod, but the tripod is a BIG benefit. The tripod will also allow you to use a longer lens offering more details. Without a tripod you will have a larger area you have to crop out thus requiring a shorter lens. The longer lens reveals more details.

Camera settings should include stopping down to an f-stop about 2-3 stops off wide open. This should minimize vignetting and deliver the sharpest details. Vignetting often shows up poorly in skies and makes blending difficult. Even if you plan to eliminate the sky, a well blended sky is easier to eliminate, trust me on that one.

For shots where you are simply dropping in a sky I find that Photoshop’s sky replacement tool is pretty good. Go to edit and select “sky replacement” which opens a tool and drops in a preselected sky image. You can select from a series of different skies or you can add your own sky image. There is a pull down menu off the sky image in the window. The other settings determine how aggressive the blending is between the AI perceived fore ground and the sky.

The construction cranes wreak havoc on the the AI so take note of that. Shifting the edge will move the general line of delineation up or down. Fading the edge will control how harsh that line is. You can tweak color temp and brightness to try and matchup the sky to the foreground and even add lighting effects. You have to play with these to get it right.

Sometimes you have to just have the AI select the sky and then you may have to tweak the selection a bit before masking it out to transparency. That is a little more delicate and time consuming but yields the most precise results.

I know this is a quick tutorial but it isn’t as hard as it may seem. Just remember to make a backup of your originals before you begin editing. You can start making everyday images a little more exciting and vibrant even if the natural world wasn’t;t cooperating with you on photo day.

Read Full Post »