Inova book- pixel perfect images ?

[lejazzcat]

Im reading this great ebook and am learning alot from it, although im now confused ,and i wonder if anyone can explain the stuff about 'pixel perfect' image size on 4-9?

It seems that hes saying that S is best for file size / image quality for the D70 CCD ?

If so, as a compromise, is it better to shoot at say M and interpolate up if you need more size, than to shoot at L?

TIA.

[Killakoala]

For the benefit of those who dob't have the book, the page is a bout the difference between LARGE, MEDIUM and SMALL when selecting image size on the D70.

Hmm, good question. The book actually explains it quite well but i understand what you are asking.

As he says, the small image is basically a compressed large image. All the information in the large image is passed down to the small image but rather than have 4 bits of information per pixel (3008x2000), there is 1 bit of information on 1 pixel, (1504x1000) but it's the same information that's on the 4 bit pixel. If you get my meaning.

However, if you enlarge that 1 bit image (1504x1000) you will have a pixel that is say, 2 times the size (3008x2000) but still only with the 1 super bit (not 4), so you lose quality and gain pixellation as each pixel is now twice the size it was before but with the same amount of tonal and colour information.

Does that make sense?

[lejazzcat]

For the benefit of those who dob't have the book, the page is a bout the difference between LARGE, MEDIUM and SMALL when selecting image size on the D70.

As he says, the small image is basically a compressed large image. All the information in the large image is passed down to the small image but rather than have 4 bits of information per pixel (3008x2000), there is 1 bit of information on 1 pixel, (1504x1000) but it's the same information that's on the 4 bit pixel. If you get my meaning.

well sort of - he calls it a super pixel - "pixel perfect"

However, if you enlarge that 1 bit image (1504x1000) you will have a pixel that is say, 2 times the size (3008x2000) but still only with the 1 super bit (not 4), so you lose quality and gain pixellation as each pixel is now twice the size it was before but with the same amount of tonal and colour information.

Ok - so were better off having the full 4 bit pixel rather than 2 x 3 bit pixels. Right

Does that make sense?

It does - but if a S size creates a super pixel(perfect- but is a little too small for printing)- does the M size also improve the "quality/detail" of the pixel over the full size L, without compromising the image size(not filesize) too much ?

So basically - is a L size image(larger file size) scaled down to M size in PS, a better image than a image taken at M? Or are they the same?

I think i understand it - it seems obvious, but i want to make sure im not missing something .

ty Killa
[/quote]

[Onyx]

It's to do with the CCD and Bayer pattern - probably mentioned in the section preceeding the one in question.

The individual bits that collect light (photosites), elements composing of the "sensor" have a coloured filter before each of them. They are arranged as green red green red, and blue green blue green etc on subsequent rows. The photosites' captured data corresponds to how much light has hit it. The camera's processor then translates this 'amount of light' value and interpolate what colour to render the pixel, taking into account firstly the colour of the filter in front of the photosite, the light values of adjacent photosites with the same and different coloured filter in front, and the position of the individual photosite with respect to the image frame.

If you shot in S filesize, the camera's processor is taking a group of 4 photosites (1 red, 1 blue and 2 green) and outputting only 1 pixel from the four, it's not the average value of the 4 but rather the predominant colour. This is evident in the sharper more defined edges you can observe viewing the image at 100%, hence the "pixel perfect" nomenclature.