Autofocus Problem
Locked
 
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
While we are currently working out suitable autofocus modes and parameters for our project we found that the way to calculate focus values by dividing mean edge intensity by mean intensity of the image - a method described in papers, widely used in the field and optionally utilized in the OughtaFocus - can lead to focus failures. Focus values of dark images calculated in this way exceed the values from in-focus images. To give an (artificial) example I have created a IJ image stack of one image (DAPI stained cells). Image 1 is the original. The images 2 to 5 or smoothed with a Gaussian (radius 0.5, 1.0, 2.0, 4.0) to simulated out of focus z-positions. To simulated the problem I have divided image 5 by a value of 20. (Cells can be visualized be increasing the contrast.) With a short IJ macro I calculated the edge based sharpness in the way the OughtyFocus is doing it: double meanIntensity = ip.getStatistics().mean; ip.findEdges(); double meanEdge = ip.getStatistics().mean; return meanEdge/meanIntensity; The values for the example stack are: Focus Value 'Edge' (slice 1) =0.396118778135568 Focus Value 'Edge' (slice 2) =0.3956836609095465 Focus Value 'Edge' (slice 3) =0.39129756958318224 Focus Value 'Edge' (slice 4) =0.3784508143789083 Focus Value 'Edge' (slice 5) =1.818627450980392 Even if in the most practical situations the method will work, I think it is worth to know about this aspect and to avoid circumstance in which this problem may become relevant. Regards, Peter ps: Attached you find a resized version of the test stack and the IJ plugin. pps: Black images (as a result of camera problems) lead to a division by zero. On 27.10.2011 20:45, Arthur D. Edelstein wrote: Hi Conor,For the "proper" experiment, I wanted to use a narrower search range for autofocus, to get capture times down as much as possible. Perhaps I was over keen with this. I have been running it for an hour or so and many of my points have already lost focus, but a few stalwarts are staying in focus, which I'm very pleased about.You may find that the OughtaFocus search time is not too badly increased if you, say, double the range, because the algorithm (Brent's method) usually converges better than quadratically. To speed up the autofocus search, you can also try increasing the tolerance so that OughtaFocus doesn't spend extra iterations over-optimizing the position. If your camera is a bottleneck, you can try imposing a crop factor or shortening the exposure time. And of course you may not need to run autofocus at every time point. I'm using a ProgRes MF. If you know of any issues with that, please let meknow and I will get in touch with the guys from JenOptik.I'm not aware of any. You can check to see if whether after you run live mode, pressing snap gives you a fresh image (and not the last image from live). Best regards, Arthur------------------------------------------------------------------------------ The demand for IT networking professionals continues to grow, and the demand for specialized networking skills is growing even more rapidly. Take a complimentary Learning@Cisco Self-Assessment and learn about Cisco certifications, training, and career opportunities. http://p.sf.net/sfu/cisco-dev2dev_______________________________________________ micro-manager-general mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/micro-manager-general ------------------------------------------------------------------------------ The demand for IT networking professionals continues to grow, and the demand for specialized networking skills is growing even more rapidly. Take a complimentary Learning@Cisco Self-Assessment and learn about Cisco certifications, training, and career opportunities. http://p.sf.net/sfu/cisco-dev2dev _______________________________________________ micro-manager-general mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/micro-manager-general |
AutoFocusTest.7z (28K) Loading...
| Powered by Nabble | See how NAML generates this page |
