May 25 2006

Fuji Super CCD – interpolation or not ?

Mathematically speaking no, but the debate continues. The Fujifilm S1 Pro, announced in Jan. 2000, was the camera which really started the debate about the Fuji Super CCD interpolation performed by the sensor and the associated electronics.

It featured a 3.07 Mpx Super CCD which has an output of 6.13 Mpx and of course, almost everyone cried interpolation.

The Fuji Super CCD has the “pixels” oriented at 45° to accomplish several things:

Better resolution both horizontally and vertically, at the expense of diagonal resolution.

Serious research shown, long before the Super CCD appeared, that the human eye is more sensitive to resolution in both horizontal and vertical orientations simply because our surrounding reality is built this way. It has something to do with the direction of the gravity but the theory is a bit too complicated to be presented in this article.

The standard sensors, both CCD and CMOS have the pixels oriented in rows and columns which offer the best resolution on the diagonals where is not really needed, both by most subjects and the human eye.

So the “Fuji Super CCD interpolation” real or not, may have a good point here. I will elaborate it a bit later.

Better sensitivity and signal/noise ratio. This is achieved by the fact that due to this diagonal orientation of the pixels and the octagonal (not hexagonal) shape of the light gathering elements, the “active” surface of the sensor is larger compared to a conventional layout. Bigger pixels mean better sensitivity and less noise.

There is no doubt that the Fuji DSLRs, including the latest Fuji S3 Pro, have wonderful noise characteristics including at high ISO.

Again, the “Fuji Super CCD interpolation” shows an advantage.

Now, let’s get back to the main debate: the increased resolution effect of the Fuji Super CCD interpolation and the actual resolution of the Super CCD sensors.

Mathematically speaking, to describe the output of a diagonal sensor array you need two times more “standard” pixels.

For a very small scale demonstration, just draw a 2×2 pixels layout on a piece of paper. If you keep it in the standard layout you have 2 pixels on both axis so 2×2=4.

If you rotate the drawing 45°, going diagonal, you will see that you need 3 pixels on both axis so 9 pixels in total to describe the output of just 4 pixels.

The result is not very precise, 9/4 is not 2 but close. This is due to the extremely small number of “pixels” used in our drawing but for bigger numbers the result is exactly 2.

So a 6 Mpx diagonal array need a 12 Mpx file to describe the output without throwing away any information.

But does a 6 Mpx Fuji Super CCD interpolation offer genuine 12 Mpx resolution ?

The answer is no. But is offers measurably better resolution than any 6 Mpx conventional sensor. But how much ? Many reviews say 8 Mpx, other 9 Mpx.

Let’s say that the result is very close to the number of actual pixels multiplied with the square root of 2, so for a 6 Mpx Super CCD the resolving power approaches that of a 8.46 Mpx conventional sensor.

Conclusion of the Fuji Super CCD interpolation debate:

The Fuji Super CCD does not literally interpolate, i.e. it doesn’t invent pixels but need the double number of pixels to correctly describe the output.

Meantime, the actual resolving power is about 1.4 x number of real pixels.

This article does not take into consideration the Bayer color interpolation which is performed in all current sensors except the Foveon design.

Related Posts:

Leave a Reply