E-M1.2 CAF: Focus Prediction Appears to be Based on Previously Exposed Images

Started 1 month ago | Discussions thread
ForumParentFirstPrevious
Flat view
drj3 Veteran Member • Posts: 7,846
E-M1.2 CAF: Focus Prediction Appears to be Based on Previously Exposed Images
25

For those who want to skip the explanation my brief conclusions first:

For moving targets, use high frame rates, turn AF Priority C = ON, don’t bother waiting for the camera to focus once the target is visible in the EVF, always shoot bursts of at least 6 images, longer bursts should give a higher percentage of focused images.

There will always be a delay in time between focus and exposure for all cameras (mirrorless and DSLR) which blackout the VF to record the image (a global shutter could solve this problem – I do not know for the Sony A9). If a target quickly changes the distance from the camera (a swallow flying toward you), the difference in time between focus and exposure will result in the target no longer being in the focus DOF for the exposure. The camera must use some type of prediction based on previous movement as well as the last actual focus distance to estimate best focus for the next image. This is not a problem for targets that move parallel to the camera where the distance to the target does not change or for slow moving targets with a large DOF.

There is also a time difference between correct focus in Live View and focus confirmation. This time delay becomes an issue when one sets AF Priority C = Off, since shutter release for the first image will be delayed until the camera confirms focus. The focus confirmation box seen in the EVF, when reviewing images of fast moving targets, will rarely be of any value, since the box represents the point in the frame where the camera focused in LV and not where the target is located in the frame of the exposed image.

I started a simple project to determine how to best photograph fast flying swallows using a Dot Sight. I quickly became interested in how well the Olympus E-M1.2 (firmware 2.0) predicted focus in CAF mode with different types of target movement. This evaluation was possible, since the swallows were flocking at the local refuge making it possible to follow the flights of many swallows with the same general approach flight pattern over several days (over 15,000 images). All my interpretations are based on placing the dot of the sight on the bird and simply pressing the shutter without concern for pre-focus accuracy. I evaluated flight sequences of 1 – 3 seconds. C-AF Sensitivity was set to +2 for all sequences. I used fps rates of 6 and 15 with the Electronic shutter and 6 fps with the Mechanical shutter. I compared sequences with AF Priority C = On vs Off. The Electronic AF Limiter was set to 15-170 feet. I could not evaluate CAF vs CAF+TR, since I was using the Dot Sight and could not see if the camera focused on the target. All data and interpretations are based on the EM-1.2 + MC14 + 300mm f4 (firmware 1.4)photographing flying swallows. I would expect similar results for other lenses, but the number of images required for acceptable focus may be different when using a shorter focal length lens with a greater depth of field or for lenses which focus at a different rate.

Conclusions:

The E-M1.2 appeared to use previous images to predict focus. The number of images required for correct focus appeared to be largely independent of the frame rate. It took approximately the same number of images to achieve focus at 15 fps and 6 fps, however, 15 fps gave a larger number of correctly focused images in the same interval of time with a higher percentage of well focused images. Higher frame rates also allowed faster (time measure – same number of images for correction) correction of focus when there was a momentary loss of focus and thus giving more focused images. However, there was a much larger number of images requiring evaluation for long sequences at the higher frame rate.

AF Priority C = Off (shutter release requires focus confirmation) reduced the number of out of focus images when the target was moving parallel to the camera and for situations with small differences in focus distance between frames. However, Off also reduced the number of images with good focus, due to the delay of the required focus confirmation.

AF Priority C = Off had little effect on number of out of focus images when the target was changing distance, but it greatly reduced the number of well focused images. The finding that requiring focus of the first image did not reduce the number of images needed to achieve good focus below the 3-5 needed at both 6 and 15 fps rates with unfocused first images, leads me to conclude that good focus prediction depends primarily upon number of exposed images. In general, more well focus images for moving targets will be obtained with AF Priority C = ON.

If the target was moving parallel to the camera and the first image was sharp, then focus was typically good for all images unless the target changed distance from the camera. If the target was moving toward or away from the camera, focus accuracy of the first image was of minimal importance, since the camera appears to need a series of images to predict focus of future images. Even if the first image was in perfect focus, the second and third images were almost always blurred. The number of initial images needed for the camera to predict movement for unfocused targets was generally 3-5 images when the target changed distance in a consistent manner. If the target continually changed flight speed and direction, then even short sequences of well focused images were rare. In the future I will generally shoot bursts 6 images or more for moving targets.

When the target suddenly changed its flight pattern, there was a loss of focus (typically a couple of frames for small deviations) until the camera reestablished a pattern to predict movement.

I thought that the Mechanical shutter might have a shorter blackout duration (1/250 vs 1/60 scan time), but I saw no difference in performance between the Mechanical and Electronic shutters. There may be some differences that would have been observed with more sensitive measures of focus accuracy of the out of focus frames or for different targets.

I would like to have been able to compare different CAF sensitivity settings, but the weather changed and the birds changed their flight patterns making any meaningful comparison impossible. The difference between the CAF sensitivity settings may be based on the number or pattern of previous frames used for prediction (requiring more frames or limiting the effect of the last frames would limit amount of change in focus distance).

I would very much like to know if ProCapture Low (PCL) images are used for focus prediction and if PCL focus performance is equal to the Electronic Shutter Low sequential setting. If the above are true, one could use PCL as a means to lower the number of initial out of focus images. For example, you could set PCL to 15 fps and the Pre Shutter frames to 10 and half press the shutter for about one second and then fully press the shutter. This would eliminate the first 3-5 frames which are often out of focus for targets which change distance quickly. If the target was moving parallel to the camera, you could just fully press the shutter instead of using the one second delay. A brief trial did not appear to support PCL focus prediction, but there were too few attempts to draw any conclusions.

See the attached image as an example of a bird changing flight pattern. The swallow was 75 feet from the camera with a DOF of 19.4. However, if focus was on the bird’s body, the bird would only need to move 9.8 inches (depth of field behind target) from the predicted position for the camera to fail to focus correctly. In the first image (#19) the wing extended toward the camera is in best focus so the body focus is okay. In the second image (#20) the head/body are no longer in acceptable focus as the bird changes flight speed/direction to catch an insect. The third image (#21) is completely out of focus, but the last image (#22) is back to acceptable focus (acceptable focus for a 5.5 inch bird at 75 feet). These images were at 15 fps rate. The bird had previously changed its flight pattern twice in 1.27 seconds. At 6 fps, it is unlikely that any images would have been in focus.

-- hide signature --

drj3

 drj3's gear list:drj3's gear list
Olympus E-510 Olympus E-5 Olympus E-M1 Olympus OM-D E-M10 Olympus E-M1 II +7 more
Olympus E-M1
If you believe there are incorrect tags, please send us this post using our feedback form.
ForumParentFirstPrevious
Flat view
Post (hide subjects) Posted by
(unknown member)
(unknown member)
(unknown member)
GBC
(unknown member)
GBC
(unknown member)
(unknown member)
(unknown member)
(unknown member)
(unknown member)
(unknown member)
(unknown member)
(unknown member)
ForumParentFirstPrevious
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark MMy threads
Color scheme? Blue / Yellow