The D70 is known to have a weak anti-aliasing filter. This theoretically improves sharpness (the uber-expensive Leica M9 doesn’t have one at all), but can cause problems with patterned subjects like fabrics as you’ve discovered.
Good eye!!!
We had a similar problem in college band. Our uniforms for the basketball games were always whimsical, and one year we wore striped referee shirts. Which generated Moire patterns whenever a TV camera was pointed at us. The solution to which, unfortunately, was that the cameramen decided to not get any shots of the band for the rest of the season.
Uh, may I ask: Isn’t this illusion in the eye of the beholder of the picture, and NOT actually in the picture itself?
Not a chance. It is definitely IN the color information in Photoshop. Move the eyedropper over the blue blob and you will select the color blue.
No, this is a flaw in the sampling of the original scene by the camera. Imagine a 1d slice of four pixels, each 3 subpixels wide: 111222333444
Now, each subpixel samples some spectral frequency distribution. For simplicity’s sake, let’s say there are 3 frequency distributions, red ®, green (G), and blue (B): RGBRGBRGBRGB
Now imagine a high-frequency black (K) and white (W) pattern focused on those twelve subpixels: KWWKKWKKWKWW
Such a pattern will then make the four pixels two colors. Pixels 1 and 4 will be yellow (min red, max green, max blue) and pixels two and three will be blue (min red, min green, max blue)
What software did you use to convert from the raw NEF file?
Moire in general is a problem with the D70 and D70s for the reasons explained above. But I’m pretty sure the the specific blue and yellow effect is a characteristic the JPEG encoding algorithm used in the camera itself. It’s possible Nikon used that same algorithm in the PC software distributed with the camera (I never used it so I’m not sure).
If you’re using an old version of the Nikon software, or the raw conversion plugins provided by Nikon, it’s likely they are exaggerating the problem. Try converting it again with some other software, and I expect you’ll see better results (not perfect, but easier to deal with).
I’m not using any Nikon software, just Photoshop CS3.
Photoshop used to require a plugin provided by Nikon to do its raw conversion. I don’t use it so I’m not sure what the deal is now, but I wouldn’t count on CS3 eliminating Nikon software.
You could try Irfanview - it’s free and I believe it has native NEF support. If nothing else it’d rule out the JPEG encoder as the source of the problem.
Oh, that’s a better possibility than my idea above. JPEG converts the image into a YCbCr colorspace (luminance, yellow - blue, green - red), does a discrete cosine transform on the three channels for each block of pixels, and then quantizes the values for each channel. A poor pick of quantization for the Cb channel could explain why all of the color fluctuations in the image go from yellow to blue.