rescue severe OUT OF FOCUS SD videos !!

[Jim Arco]

As far as I can tell, there is no tool (yet) to truly fix out-of-focus images.

There are several plug-ins available claiming to fix the problem, but they mostly provide a combination of several sharpening techniques. What they do, they do very well and with some out-of-focus images, they can work wonders. Nik and Topaz come to mind, but I haven’t used the trials in while, so it may be other vendors. Pretty sure they are Photoshop plugs (only), so you’d have to pass your video through PS.

I took your image and tweaked the shadow-highlights, then some unsharp mask with a low amount (20%) and a high radius (7 pix) to get a little improvement.

One problem you will have with many video sources (particularly SD) is the large amount of edge-enhancement added by most video cameras. Any significant amount of sharpening exaggerates the edge-enhancement, making the image look “cartoon-y.”

If you find something to make big changes in out-of-focus materials, I’d love to hear about it.

Jim
ColorBurst Video

[Jim Arco]

I seem to be having trouble uploading the edited image. Here is another attempt.

[Nick Ag]

Hi Jim, did you copy to your post the embed code or link that it shows next to the upload images panel ? I guess its the only way to add images here
Thanks for your valuable effort by the way

[Andrey Sibiryakov]

Hi Rig.

Some days ago I’ve run across a program called Perfect resize pro, that uses the method of enlargement different from others.
I was impressed by results.

You can export to sequence,
enlarge every frame x2-3 times,
than import back to SD.

That will work.

[Jim Arco]

Ooopss! (One more time.)

[Chris Wright]

first off, you need to interpret fields on. There’s scanlines in there.

secondly, I wrote a quick template that uses find edges as an alpha matte for unsharp so it doesn’t go crazy. then I added a deartifactor in there because the chroma is artifacty. Is that a word? As for SD to HD, bicubic resizing is the best. You get it from photoshop extended.

ae sharpen 5.5 aep
https://www.mediafire.com/?itsho533td3843c

[Nick Ag]

Thank you Chris, I really need your guide, I’m so excited to try your method, but I’m not exactly an expert, when you mean “interpret fields on” do I do that on “interprate footage” then “Fields and Pulldown” (separate fields ?) or where exactly ?

unfortunately the template you made was done in AE CS5.5 and I only have the CS4 but I guess I will try to get the CS6 trial, and finally do you mean that I should resize the video on photoshop extended but why exactly ? I mean, increase the size and then reduce it again to SD while editing ? SD is perfect for the customer but as long as the out of focus is improved, this out of focus is severe, if you can take a look to my first post there is a frame sample in SD
Thanks for your valuable help Chris, I hope you can guide me a little bit, I’m really desperated !!

[Nick Ag]

Ups, seems like I won’t be able to open your ae template, I got 32bit windows and CS5 and UP require 64bit windows, I just have one computer and I can hardly update to 64 bit at this point

I also found this method (jpg attached), but seems to be too advanced for me, I can not understand some steps

[Ht Davis]

I know it’s an old thread. Thanks for the info, Nick Ag. I’ll have a look at that and get back on what to do. I’m a novice with AE, but not photoshop. I’ll try to explain the above guide for you. It has a very Photoshop feel.

Basically, it wants you to dupelicate the CHANNEL layers in LAB mode. You’ll have Luminance channel instead of RGB. You’ll dupe this 3 times: shadow, highlight, midtone, and then remove areas from each channel that don’t apply. Now you can feather or blur the edge of these to allow the transition areas to be mixed. If there are specific or special areas that need to be tracked\traced, make another copy for that item (3 copies actually, since you’ll be applying this masking separately, and they will only contain the information from that object or person for highlight, midtone, shadow). Now you can apply a high-pass or edge detection filter to the masks (this is the feathering from before, but you can blur the edges further if you want). Once that is complete, you add sharpen filters to each mask to increase the noise for each channel, and check in between to see the result before attempting it again. In photoshop, you’d do this in layers. Unfortunately, there isn’t any counterpart in AE.

Here’s another AE way that actually came out of a photoshop technique (WARNING it also uses photoshop and a lot of scratch space):
First, create a comp by dropping footage onto the button or right click the clip and choose create comp from selection. This will give you a comp to work with. Now drop the video file onto the comp so there are two copies of your video. Use the INVERT effect on the top layer to reverse it’s data to a photonegative. Set the blend mode to color dodge or linear dodge, you’ll get pure white. Apply a Guassian Blur and adjust until you see mostly edge detail. Now we have options. We can go simple and just let the midtone\shadow areas go smooth, while we tighten up the edges a bit, or we can work on the midtones a bit. Repeat the above procedure, but take the blur to a point where some of the midtone data is there. Using these two precomps, we’ll need a third, so you can actually just use one of them to create it. Put the edge precomp (the first one) layer on top of the extended edge\midtone precomp. Invert the edge layer, like before, and set it to Color Dodge or Linear Dodge for a blend mode. This comp is now the Midtones only.
Now we can use the edge and midtone precomps to do two things. We can operate on the edges and midtones to apply noise or affect the transition weight (as you get closer to the edge of something, the light transition goes to an almost black, then alters to the new hue from there outward; the weight or speed of this transition to black and how far black it goes can be indicative of the distances between objects, and it is often referred to as microcontrast). If you choose to affect the smoother areas of the midtones, you may want to simply add noise. This can be done easily enough. Since you already have the base noise in the precomp, you can sharpen this precomp as a video layer in another comp, or just add some random noise, but use a a light touch, as too much noise will distort your image rather than improve it. You could just as easily run a darken operation (a burn of the shadows) on the layer, and use that to bring out areas of your midtones rather than adding noise, or do both. Once they are done, you drop them on a main comp with your video layer underneath, create a shape (for your main mask), and apply a blend mode. I would go with black as your shape color, because you can then Paint in the areas you want to affect most. Using the mask will also let you rotoscope quickly, so you can mask out the correct areas and make only a few adjustments. Now use the midtone layer in the mask, and you should see the midtones get more noisy, using a luminance, or darken type of blend, which will make minute details show more clearly. You can use the opacity of the midtone precomp to decide how strong or weak the effect should be. You can do the same with the edge detail. I’d go ahead and use the sharpen tool or the unsharp masking tool. These will both work close in on the main transition area. However, you are best off to go one more step, and make this a Black and white, so create another comp and drop the edge detail comp into it, use the B&W effect and now you can use this new comp to adjust the transition more normally. Repeat what you did with the midtones. You can apply more noise or sharpening, but try to keep the edges from thickening too much. We want some dissolve areas. So, there is an easy way to do this. You can make the lines thick, then apply a blur to feather them until only the main edge is all black. Now use this precomp in the comp that has the midtone adjustment and the video layer (your main output comp), apply it using a shape mask like before.
If you prefer the photoshop path, you can set up your edge and midtone details, then output those to image folders for photoshop to batch process. If there are common areas in the images (IE the camera doesn’t zoom or move), you should be able to mask out areas by simply erasing them in an action script and applying that script to entire sets of images, saving them back out to be overlayed on top of your video. Using the same blending mode and shape technique, you can apply your adjustment. To sharpen the edge, you darken the microcontrast, to soften it, you brighten up the microcontrast.