Canoscan 9000F vs Minolta DSMP

I’ve recently started using a Canon Canoscan 9000F to speed up my scanning workflow, using it primarily to make fast, low resolution quick’n’dirty index scans. I’ll write more about that in another post, but I thought it would be interesting to see how the Canoscan stacks up against the Minolta Dimage Scan Multi Pro in the overall image quality stakes. I should emphasise at the outset that I do not expect a $350 flatbed scanner to match a $3500 film scanner, but you never know - and since the film scanner is no spring chicken, I have to start thinking about what I’m going to do when the day comes that it stops working.

Comparing scanners can be fraught with peril, as my colleague Bruce Percy discovered not so long ago, but I’m going to leap in regardless.

So, I took a recent a Velvia 100 XPan shot of a local cow I happened to pass by the other day, I scanned it on both scanners, in both cases using SilverFast AI Studio, both with Auto-IT8 calibration, and to 48-bit HDR Color with Digital ICE enabled on the Minolta, and 64-bit HDRi on the Canoscan (Silverfast does not yet support HDRi on the Minolta).  In both cases I used Silverfast’s Multiexposure, and scanned at 4800dpi. This is the optical resolution of the Minolta; the Canoscan claims 9600dpi, which seems ... unrealistic ... and in any case, various authorities claim that 3200 dpi is as much as you can extract from 35mm film. However, in my experience, the Minolta does deliver a little more data at 4800. I also used the Scanhancer on the Minolta, to get the best possible result.

I made the scans,then opened both in Silverfast HDR Studio. Here, for the sake of a level playing field, I used the Auto function to adjust tonal balance, and output to ProPhoto RGB colour space.

I was interested in three factors: dynamic range, colour fidelity, and detail. The first thing I looked at were the histograms from the two scanners. First Canon, then Minolta:

Histogram from the Canon scan

Histogram from the Minolta scan

It is fairly evident from these that the Minolta scan holds a greater range of data values. There is also an interesting offset in the bright peak in the blue channel in the Canon’s histogram.

So, now to the colour comparison - and introducing the aforementioned cow: first the Canoscan, then the Minolta

Well, we’re not too far out. The blue peak in the Canon histogram shows itself in the blue tint to the sky, which arguably is more attractive, however, and you’re going to have to take my word for this, the Minolta version is pretty much spot on compared to the Velvia transparency.

I’m not entirely sure what is going on here, as I would expect the Auto-IT8 calibration to smooth this out. However, I’m not sure that using a 35mm IT8 target on the Canoscan is ideal (I need to check with Lasersoft), and of course this is just one image ... hardly a trend!

Of course I’m not too concerned about this: the Minolta is my reference scanner, and the scans that Silverfast delivers from that are extremely accurate. However, my general and possibly naive assumption is that while dynamic range and resolution will of course vary, in a fully calibrated system, and within device limits, the colour gradation should be quite close.

The last thing to look at is detail. In both cases I applied light “capture” sharpening using PixelGenius Sharpener (in future I may evaluate using Silverfast USM sharpening for this step). Below are 100% crops from each scanner, Canoscan first. Please do note that JPEG compression does degrade things a little.

Here the difference is quite clear. From other experiments I’ve observed that the Canoscan’s effective optical resolution peaks at around 2400dpi - there is no difference in scan quality at higher resolutions. The Minolta is resolving grain (what little there is in Velvia 100) and this has been slightly accentuated by the light sharpening.

In terms of usable detail, however, there’s not a lot in it, and both will print fine up to A3. What is more important is that the Minolta resolves considerably more shadow detail. I suspect that the Canon is losing out by not having any control over focus. In fact here I let the Minolta auto-focus, but for critical cases, I often use manual focus, which can make quite a difference (it’s a very slow process though). Here, actually, I thing the AF has done a pretty good job.

So, my conclusion is that the Canoscan delivers a very commendable performance, especially considering the price. I’m a bit intrigued by the inaccuracies in the colour gradation, and I’ll probably investigate that a little further when I have time - but I don’t consider it a big deal.