|
|||||||||
|
Thread Tools | Search this Thread |
September 3rd, 2009, 12:07 AM | #1 |
Trustee
Join Date: Mar 2006
Location: Boise, Idaho
Posts: 1,997
|
Identify this image problem?
The attached image shows several problems upon closer examination (or even not so close!).
Problems are: 1) Barrel distortion (use of WD58 zoomed all the way out in a small room?) 2) Discoloration on edges of white objects (with a green smeared above and blue smeared to the right) So my questions are 1) What is this color problem called? 2) Is it a bad sensor problem? A WD58 lens problem? A camera design flaw? A miss-aligned prism / sensor / both? This was shot with the WD58 mounted directly to the cam (no UV filter in between). Also attached is the full size image for comparison. |
September 7th, 2009, 12:52 PM | #2 |
Trustee
Join Date: Mar 2006
Location: Boise, Idaho
Posts: 1,997
|
Anyone have any insight?
|
September 8th, 2009, 07:38 PM | #3 |
Major Player
Join Date: Sep 2006
Location: Centreville, Maryland
Posts: 258
|
My guess would be chromatic aberration. Chromatic aberration - Wikipedia, the free encyclopedia
__________________
http://twitter.com/tedmcneil |
September 9th, 2009, 03:57 PM | #4 |
Wrangler
Join Date: Sep 2001
Location: Northern VA
Posts: 4,488
|
Barrel distortion is common to very wide angle lenses, some more so than others. It is most noticeable when there are clear vertical edges and clean backgrounds such as the cabinet edge. In short, not unexpected.
The color fringing may be an artifact of compression/decompression, and perhaps some image processing as well, noticeable because the the scene above and to the left are flat with no texture, color/contrast variation or detail. Probably not noticeable in "average" scenes on typical SD TV sets The first image size is 655x480 pixels, a somewhat unusual size. Was it a full frame grab?
__________________
dpalomaki@dspalomaki.com |
September 9th, 2009, 04:04 PM | #5 | |
Trustee
Join Date: Mar 2006
Location: Boise, Idaho
Posts: 1,997
|
Quote:
|
|
September 9th, 2009, 04:06 PM | #6 |
Trustee
Join Date: Mar 2006
Location: Boise, Idaho
Posts: 1,997
|
Regarding the color fringing, is this a result of miss-aligned CCDs or the prism? Or something that can be fixed or is it more a property inherent to the GL2 line and thus something that must be lived with?
|
September 11th, 2009, 04:48 PM | #7 |
Wrangler
Join Date: Sep 2001
Location: Northern VA
Posts: 4,488
|
The fringing you see is a combined effect of JPG compression at a boundary condition (the sharp contrasty edge) and a slight difference (about half a pixel width) in the registration of the green CCD compared to the red & blue CCDs. This accounts for the greenish tinge on the left and top edges/diagonals, and the magenta tinge on the right and bottom edges. I can reproduce the effect easily in photoshop.
The JPG compression results in the smear effect, the pixel registration for the color. This pixel registration might be intentional as an artifact of pixel shift, used by canon in some 3 CCD camcorders (XL1 and GL1, not sure about the GL2) to improve resolution in frame mode and in digital zoom modes. At the CCCD pixel level it amounts something on the order of 0.0003 inches. It was not an much of an issue with SD viewing platforms, but becomes apparent at the pixel level and when viewed on HD systems and higher compression files.
__________________
dpalomaki@dspalomaki.com |
September 11th, 2009, 04:50 PM | #8 |
Trustee
Join Date: Mar 2006
Location: Boise, Idaho
Posts: 1,997
|
I see. so the issue isn't exactly a hardware problem, so much as a part of the design of the camera (hence unavoidable and not something that can be fixed).
|
September 12th, 2009, 01:18 PM | #9 |
Wrangler
Join Date: Sep 2001
Location: Northern VA
Posts: 4,488
|
That would be my take on it.
I've attached two GIF files to show this, created in Photoshop. They illustrate the effects of scaling, diagonal edges, and JPG compression. While not a precise analog of your scene or the camcorder they are illustrative. The first is from a basic image created at 1280x960 pixels. The green and magenta borders were 1 pixel wide and will correspond to 1/2 pixel width when scaled to 640x480 resolution. Note the clean crisp border. The second is a from a JPG created by down scaling the original graphic to 640x480, rotating it 10 degrees, and it saving as a high quality JPG file. Both images were captured from a Photoshop zoom-in so you can clearly see the individual pixels. I used GIF to avoid additional compression artifacts. Also this helps illusrate why HDV looks so much better on a HD display. In images with complex natural content (not flat fields) the edge artifacts are much less apparent because they are masked by the natural variations in the image.
__________________
dpalomaki@dspalomaki.com |
September 13th, 2009, 03:08 PM | #10 |
Trustee
Join Date: Mar 2006
Location: Boise, Idaho
Posts: 1,997
|
So the follow on is "is this image problem inherent to all 3 chip cameras?" or jsut to the canon line due to their "pixel shift" (which I had heard about prior to purchasing the camera). Or is this problem a GL2 thing and not present in XL1/XL2/XHA1, etc?
|
September 14th, 2009, 04:19 AM | #11 |
Wrangler
Join Date: Sep 2001
Location: Northern VA
Posts: 4,488
|
The edge blurring effects would be common to any imaging system that uses compression. How much would depend on the compression scheme used.
The color/tint issue would depend on the specifics of the image prior to comprtession, and that would require some testing. Note that the color mask scheme used on one-CCD camcorders would like produce color fringing as well. One would have to run tests with the various camcorder to find out. Keep in mind that all imaging systems are compromises involving cost, quality, and the limits of human perception in typical use. Consider the NTSC system of broadcast as a classic example. The GL2 was designed with SD in mind, the XHA1 with HD in mind.
__________________
dpalomaki@dspalomaki.com |
| ||||||
|
|