Go Back   Steve's Digicams Forums >

LinkBack Thread Tools Search this Thread
Old Oct 29, 2002, 7:01 PM   #11
Senior Member
Join Date: Aug 2002
Posts: 2,162

Yes, I understand there are other factors connected with lenses and I think I've got it now. It's no good looking at the ccd spec on it's own then, to determine what the resulting picture sharpness will be be?

So real pictures and resolution charts are a fairer way to compare digicam manufacturers performance claims.

I remember when a certain TV manufacturer produced the first frame doubling TV sets (100Hz) - Yes there's less flicker, but I'm still waiting for one that doesn't produce temporal motion artefacts! I'll put 3-6Mpix interpolation in the same bag, until somebody proves otherwise with their pictures on this Forum!

Incidentally, there's been a lot of discussion about Compact Flash and speed. I've been using Smartmedia, had no probs with speed and get no freezes in 30fps movie mode.
voxmagna is offline   Reply With Quote
Old Oct 29, 2002, 9:36 PM   #12
Join Date: Oct 2002
Posts: 58

Yea, I think you're thinking about it correctly now. When a designer makes a camera, they're concerned with overall image quality. Image quality is the goal. There are many different ways to measure the quality of an image, and each aspect of lensmaking and camera making addresses one or more of these issues. One is linearity, which is affected by spherical aberrations in the lens. Another is resolution, affected by the CCD. Relative color and absolute color distances are another. Dust and scratches. Stibbling, etc, etc.

As you pointed out, picture sharpness is another important contributor to image quality. This term too has a specific meaning, though. "Sharpness" has to do with edge differentials. In other words, if you have an image half black and half white, divided by a perfect edge, then if you take a picture of that the ideal image should show no grey values--only black pixels and white pixels. Of course, the human eye can only resolve so much detail, so over a certain density of dots per inch (around 80) there is no effective difference between the ideal image and an actual image.

Keep in mind that sharpness and focus, though, are different things as described in the tutorial on sharpening at luminouslandscapes.com. Focus describes edge boundaries, while sharpness describes edge contrast. They are related concepts but not the same.

Resolution charts emphasize resolution over the other aspects of image quality, but you can learn more about the image quality from a resolution chart than just resolution if you examine it carefully. Just as an example, though, one thing a resolution chart will tell you little about is color casts and things like that.

Also, you have to keep separate the idea of non-local aspects of image quality versus local aspects. Image sharpness, color casts, focus, and resolution are all non-local aspects of overall image quality because they each affect every region of the image in the exact same way. Spherical aberration, dust or scratches...these things affect different regions of the image differently. This is an important distinction because non-local problems affect the information contained in the image uniformly, whereas local problems usually only affect specific regions.

The biggest thing to remember is that, no matter which aspect you're looking at, if it becomes too far out of balance in terms of how it affects overall image quality with all the other aspects, the image as a whole is degraded.

Now, about interpolation. What you need to understand about this is how the super CCD works. When you view a printed image, or an image on-screen, you are essentially viewing a rectangular grid of pixels that make up the image. If the CCD that captures the image is a rectangular grid, there is a one-to-one correspondence of photocells in the CCD to pixels in the image. If you zoom in on such an image, at some point you will notice "pixellation"--that's the point at which one image pixel is magnified over a large enough number of screen or ink pixels that it can be seen by the human eye. "Interpolation" is when the ratio of image pixels to photocells in the CCD is higher than 1:1 (3:2, 2:1, etc).

Now if you picture a rectangular CCD made up of square photocells, you can imagine the distance between the centers of neighboring photocells. Now also picture the distance between centers of diagonally neighboring photocells. The ratio of these distances determines how much interpolation you can do vertically and horizontally before the diagonals in the image get too stretched. Now imagine the super CCD, which is tiled not with square photocells but hexagonal ones. The distance between centers vertically and horizontally is much higher than distances between centers diagonally, meaning that the same percentage of interpolation will be far less noticeable.

One additional point about the super CCD. There is no longer a one-to-one correspondence between photocells in the CCD and square screen or ink pixels. This means that each pixel collected on the CCD will not necessarily appear as a single pixel in the image. In other words, the super CCD actually collects information that only appears *between* photocells in a normal CCD. This is another way of understanding why a super CCD image will retain higher levels of quality under interpolation--with a normal CCD, interpolated pixels line up in horizontal and vertical rows next to one another. With the super CCD, interpolated pixels are evenly dispersed over the entire image and two interpolated ("invented") pixels will not appear next to each other until the ratio of image pixels to photocells becomes higher than 2:1.

One final factor that makes the super CCD more effective is the fact that there are rarely perfectly horizontal or vertical lines in an image. So do this experiment. Get a sheet of paper and draw a large grid on it made up of square cells. Now draw a diagonal line, at any angle, through the cells and completely color in any cell the line touches. Now do the same thing over, but this time tile the surface with hexagonal cells. Even if you make sure the area of each square cell is exactly equal to the area of each hexagonal cell, you will find that any single diagonal line touches more of the hexagonal cells than square ones...i.e., it's "higher" effective resolution because it's made up of more pixels. The only lines, in fact, that you can draw which touch more square cells are perfectly vertical and horizontal.

sever is offline   Reply With Quote

Thread Tools Search this Thread
Search this Thread:

Advanced Search

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are On
Pingbacks are On
Refbacks are On

All times are GMT -5. The time now is 12:10 PM.