Erratic Multiple Scanning and Bad Shadow Performance
PostPosted: Tue Oct 21, 2003 10:48 pm
Having an ArtixScan 4000tf, which, with its claimed dynamic range of 4.3, is supposed to be able to capture even the darkest shadows of slides. However, it is obviously not capable of that, at least not if SilverFast 6.1.0 is used. With normal slides, this may not be much of a problem, but in slides with really dark areas, there are details in these deep shadows that SilverFast is not able to capture - they are buried in a murky disgusting haze. And... an IT8 calibration worsens it...!
In trying to save some underexposed slides of Fuji Sensia 100, I have tried all kinds of counter measures, especially Multiple Scanning which is enabled with this Scanner. The result, however, is very erratic.
My question to LSI is: Are there plans of improving the multiple scanning function with this scanner, or any other improvements in order to really bring out the full range in a slide such as Fuji Sensia? Details of my tests follows:
I know that ArtixScan/Polaroid have not done much to support multiple scanning, and that they think the CCD (the "new" one in the FireWire models) is so good that they don't need multiple scanning. In practice however, at least SilverFast cannot bring out the deep shadow details. It has been said that this scanner is not mechanically sturdy enough to register multiple scanning without loss of sharpness, but nevertheless, the function is enabled in SilverFast with this scanner model, so I have tried it.
I have tried scanning a slide at x1, x2, x4, x8, x16, and then the same procedure again and then with another slide. I always scan using 48 bit 4000 dpi HDR gamma 2.0, since I don't want SilverFast to do any automatic color or contrast corrections - I want to do all correction work in PhotoShop under full control. I have it calibrated (actually characterized) using a Fuji IT8 slide from ***** (coloraid.de).
The really strange thing is that the level of background noise varies VERY much on different occasions! In one instance, the x16 is by far the best, whereas the x8 has a very high level in the blacks - more than the x2 - although the noise is very smooth; in another instance, the x8 is the good one with lots of shadow detail whereas the x16 is the one with a very high level in the blacks - often som greenish haze. Sometimes the x4 is the most quiet one! I find this very strange and very disturbing, making multiple scanning useless.
As for sharpness, it varies a lot between occasions; sometimes virtually nothing is lost, sometimes areas of the image suffers significant loss of sharpness. Most baffling is this occasional behavior: compared to a x1 scanning, it appears as if the image has been buckled into waves - shortening & extending objects!
As an alternative, I have tried to optimize the lamp brightness. Granted, if I drive the lamp Lightness way up, it really can pick up all the details in the shadows beautifully, as much as I can see using a good loupe, but at the price of severely clipped highlights. Since I want to batch scan my slides, I don't want to dive into individual optimizations of Lamp control at the scanning stage; now, checking with the IT8 slide, I can only increase the Lamp Lightness to 8 before clipping occurs in the red, and 8 is much too little to really bring out all shadow detail.
As a final step, I first looked at the different channels, and I discovered that red is the worst; and not only that, red behaves very strange, showing LOWER levels in areas which are somewhat brighter in other colors compared to areas that are supposed to be close to black!? That led me to suspect color management. So I scanned it (at x1) using color management off, i.e no IT 8 calibration and no profiles. Now, the shadow performance is better! Still noisy though, but red now behaved normally as the other channels. So obviously, using IT8 calibration quite drastically worsens shadow performance! Reducing the saturation in deep shadows helps a lot, but such tricks are lousy substitutes for the real thing - namely a good close-to-noise-free capture!
Hoping for serious comments and plans for SilverFast to squeeze out more from an ArtixScan 4000tf.
In trying to save some underexposed slides of Fuji Sensia 100, I have tried all kinds of counter measures, especially Multiple Scanning which is enabled with this Scanner. The result, however, is very erratic.
My question to LSI is: Are there plans of improving the multiple scanning function with this scanner, or any other improvements in order to really bring out the full range in a slide such as Fuji Sensia? Details of my tests follows:
I know that ArtixScan/Polaroid have not done much to support multiple scanning, and that they think the CCD (the "new" one in the FireWire models) is so good that they don't need multiple scanning. In practice however, at least SilverFast cannot bring out the deep shadow details. It has been said that this scanner is not mechanically sturdy enough to register multiple scanning without loss of sharpness, but nevertheless, the function is enabled in SilverFast with this scanner model, so I have tried it.
I have tried scanning a slide at x1, x2, x4, x8, x16, and then the same procedure again and then with another slide. I always scan using 48 bit 4000 dpi HDR gamma 2.0, since I don't want SilverFast to do any automatic color or contrast corrections - I want to do all correction work in PhotoShop under full control. I have it calibrated (actually characterized) using a Fuji IT8 slide from ***** (coloraid.de).
The really strange thing is that the level of background noise varies VERY much on different occasions! In one instance, the x16 is by far the best, whereas the x8 has a very high level in the blacks - more than the x2 - although the noise is very smooth; in another instance, the x8 is the good one with lots of shadow detail whereas the x16 is the one with a very high level in the blacks - often som greenish haze. Sometimes the x4 is the most quiet one! I find this very strange and very disturbing, making multiple scanning useless.
As for sharpness, it varies a lot between occasions; sometimes virtually nothing is lost, sometimes areas of the image suffers significant loss of sharpness. Most baffling is this occasional behavior: compared to a x1 scanning, it appears as if the image has been buckled into waves - shortening & extending objects!
As an alternative, I have tried to optimize the lamp brightness. Granted, if I drive the lamp Lightness way up, it really can pick up all the details in the shadows beautifully, as much as I can see using a good loupe, but at the price of severely clipped highlights. Since I want to batch scan my slides, I don't want to dive into individual optimizations of Lamp control at the scanning stage; now, checking with the IT8 slide, I can only increase the Lamp Lightness to 8 before clipping occurs in the red, and 8 is much too little to really bring out all shadow detail.
As a final step, I first looked at the different channels, and I discovered that red is the worst; and not only that, red behaves very strange, showing LOWER levels in areas which are somewhat brighter in other colors compared to areas that are supposed to be close to black!? That led me to suspect color management. So I scanned it (at x1) using color management off, i.e no IT 8 calibration and no profiles. Now, the shadow performance is better! Still noisy though, but red now behaved normally as the other channels. So obviously, using IT8 calibration quite drastically worsens shadow performance! Reducing the saturation in deep shadows helps a lot, but such tricks are lousy substitutes for the real thing - namely a good close-to-noise-free capture!
Hoping for serious comments and plans for SilverFast to squeeze out more from an ArtixScan 4000tf.