AJSJones Offline Upload & Sell: Off
|
All this strong feeling and the thought of trying to undo aliasing by using Gaussian blur made me go get much of the post from POTN. The effect of the 1 pixel AA filter is waaaaaaay less than a 1 pixel Gaussian blur. The loss of detail from the AA filter is actually not much - if you actually see before and after - the pictures below surprised me, especially by how much one can recover, with even simple USM, let alone potential deconvolution algorithms, with a known PSF. If you are not into tech stuff, just look at the pictures to see the effect of a 2x2 AA filter. It's definitely noticeable but not worth "hating" Try it on your own image (one that has not yet been sharpened at all - perhaps one of your favourite brick wall) (Sorry the links below don't work iPad error - the original POTN post has them)
From POTN:
I've been looking for good solid info on AA filters for a long time (rather than speculation based on "This is fuzzy it must have a strong AA" This site shows that the 30D and 5D's AA filters have thicknesses that are proportional to their pixel spacing, consistent with similar "strength" -but no indication of the pixel shift magnitude) Buried on MaxMAx's site is this interesting [B][U]direct look[/U][/B] at the effect of the AA filter on the 5D2, as well as a close-up of the CFA. I recommend a visit to that link, but here are some quotes
[QUOTE] For the Canon 5D II sensor, [B]it appears that they displace the image approximately by one pixel.[/B] The complete OLPF has two layers. These pictures show 1 layer or 1/2 of the blur filter. [B]The 2nd part blurs the image 1 pixel in the vertical direction.[/B] This means that for any one point of light, you end up with 4 points separated by 1 pixel or the same size as one R-G-G-B CFA square. You have 4 points because the 1st layer gives you 2 points, and then the 2nd layer doubles those to 4 points.
Curiously, [B]if the manufacturers wanted to completely eliminate moire, they would have to blur the signal at by a minimum of 3x3 or 9 pixels because of the R-G-G-B color filter array[/B]. In this study, the Canon 5D II has a 2x2 or 4 pixel blur filter. Our guess is that the manufacturer didn't want to blur the image quite that much because of the loss of resolution, so they picked a midpoint to stop some moire and not sacrifice too much resolution. Because of their choice for a 2x2 matrix, [B]you can see color moire patterns on a 5 D MK II[/B] though you won't see it for a monochrome image. [/QUOTE]
Canon's description of their filters shows exactly this AA filter design and it is optimal if it is 1 pixel shift in each direction. The MaxMax picture actually[B] proves[/B] this
Here are a couple of tests I did (OMG) 7 years ago, based on the understanding fron Canon's information. The hummer's eye is original pixels (not scaled, just cropped out of an RGB image). To simulate the AA filter effect I duplicated the layer, set it to 50% , shifted it 1 pixel left and flattened. I then duplicated that "double image", set it to 50%, shifted 1 pixel up and flattened. The result is the middle image. I was suprised how mild the effect was. The next image is the result of Canon's recommended start point to "undo" the AA filter's effect. Not bad, I thought. After that experiment, I stopped worrying that the AA filter was stealing gobs of detail that I could never retrieve. The picture below that is from a trip to Alcatraz (5D, f/9, 1/1000, ISO 640 350mm). I compared the "1 pixel left/one pixel up" AA filter with a 1 pixel radius Gaussian blur - ew yuk, the AA filter ain't so bad.... The USM recovers the lost feather detail quite well.
These days, I still use 300/0.3 for a quicky but for the utmost, I use Topaz "In Focus" to extract the max detail for my landscapes. YMMV As for the 5DS vs R debate, I suspect that much of the time, my overall Airy discs and other blur sources will be close to functioning as my AA filter, and Topaz In focus (or similar based on some component of deconvolution algorithm) will achieve a good result. Seems to work well on my 7D, with similar sized pixels. I await some thorough testing with the cameras when they are finalized.
|