Version 1.21 Epson raw converter

[Peter Tinson]

This is available for download from http://www.epson-europe.com/ under support, Digital Photo Device, RD-1, driver and software downloads.
Peter

 

[Gid]

Well spotted. Must download and try out tonight.

 

[Gid]

I just downloaded and tried this out. Works a treat. Haven't tried all the new features yet. The interpolation appears as part of the image processing dialogue.

Enjoy.

 

[pfogle]

Can someone please tell me what the purpose of the interpolation is? Is it just for smoother transitions on large prints?

I like the curve adjustment feature - is that new? I don't recall it from the previous version. I think I'll be using this for b/w for a while, and see how I go. Up to now I've been doing the IR pix in RSE, then conversing to mono with the channel mixer in PS, but this version of EPR looks good using the b/w green filter option. I'll try to do a comparison this week and post the results.

I notice others (particularly Sean, if I recall correctly) have preferred the b/w conversions of EPR, but personally I've never liked the interface or the controls, so I've avoided it. Maybe now's the time to give it a fair go!

 

[RML]

I downloaded it yesterday evening while converting a bunch of shots. Haven't installed it yet, though. That will have to wait until I have more time.

 

[jlw]

Can someone please tell me what the purpose of the interpolation is? Is it just for smoother transitions on large prints?

Basically, yes (although it's also useful for sending files to moronic printers who say "You won't get good quality unless it's 300 dots per inch." Use of the phrase "dots per inch" in connection with pixel dimensions is always a warning sign that the user doesn't know what he's talking about.)

Epson says that its interpolation technique retains more edge sharpness than the bicubic interpolation used in Photoshop (e.g.) That's plausible -- it should be possible to tailor an interpolation algorithm to a specific sensor array and get better results than a one-size-fits-all version -- so I'll be interested to try this and see if it proves out.

 

[lightwriter]

Basically, yes (although it's also useful for sending files to moronic printers who say "You won't get good quality unless it's 300 dots per inch." Use of the phrase "dots per inch" in connection with pixel dimensions is always a warning sign that the user doesn't know what he's talking about.)

Could you elaborate on this statement. Not sure what you really mean. Are you referring to printer dpi vs. display ppi?

Epson says that its interpolation technique retains more edge sharpness than the bicubic interpolation used in Photoshop (e.g.) That's plausible -- it should be possible to tailor an interpolation algorithm to a specific sensor array and get better results than a one-size-fits-all version -- so I'll be interested to try this and see if it proves out.

The PS bicubic interpolation is not very good at creating a lot of pixel in one pass. I rarely ever need a larger file, but when I do I have found that interpolating in steps produces far better results. If Epson created an algorithm taylored to their files, then I agree that it should produce far better results than a generic version like the one found in PS.

 

[Sean Reid]

I just did an initial review of the new software on the site and it is a big improvement, incorporating almost all of the changes we have (collectively) been asking Epson for. When time allows, I'll do a formal comparision of the Epson up-res'ing vs. other methods but the initial results I've seen from their method look very good.

Cheers,

Sean

 

[jlw]

Could you elaborate on this statement. Not sure what you really mean. Are you referring to printer dpi vs. display ppi?

I hope nobody was insulted by my crack about morons and "dots per inch" -- I realize "dots" is used casually by lots of knowledgeable people.

I was referring specificallyto working with the fringes of the printing trade (which I do every day because I'm in the graphic design business.) In this type of situation you're often exposed to people -- clients, sales reps, and the like -- who aren't really trained in the technology, but like to throw around technical terms without really understanding what they mean. Unless you size up the person's level of knowledge, you can be misled into making expensive mistakes.

Specifically, with "dots" vs. "pixels," there are two places people seem to be led astray:

-- People in the printing trade usually use "dots per inch" to refer to halftone dots. There's an important relationship between halftone dots and image pixels (at the final printed size, you need to have at least 1.4x as many pixels per inch as dots per inch to avoid the risk of moiré patterns) so when people are casual about mixing these up, it can be hard to be sure you're making the correct calculation.

-- Pixel "pitch" (in units per inch) is meaningless when talking about an image itself: pixels don't have any specific size until you print them. So when somebody tells me s/he needs a "300 dpi" image file, I don't know whether I need to supply a 300x300-pixel image that can be printed at a size of 1 inch square, or a 3600x3600-pixel image that can be printed at a size of a foot square, or what.

In both cases, graphics-trade people who really understand the process are very aware of these problems, and are careful to use the terminology precisely so as to help avoid them. When you're working with graphics-trade people who don't use the terms precisely, it's a tip-off that they may not be well-informed, and that you'll have to dig deeper to make sure you're getting the information you need to supply the right type of file.

 

[lightwriter]

-- Pixel "pitch" (in units per inch) is meaningless when talking about an image itself: pixels don't have any specific size until you print them. So when somebody tells me s/he needs a "300 dpi" image file, I don't know whether I need to supply a 300x300-pixel image that can be printed at a size of 1 inch square, or a 3600x3600-pixel image that can be printed at a size of a foot square, or what.

No offense taken. I figured you were on to something of value. Very informative. I knew about this but keep "forgeting" b/c as you said a lot of people use the terms liberaly. One final question though as am I still a bit confused. If one needs 1.4x as many pixels as dots to prevent moire and someone asks for a 300 dpi image - wouldn't you take the 3600x3600 pixel image and divide by 1.4 first and then by 300 which gives an image 8.57 inches square?

Thanks for the info. Any advise on printing with an inkjet - I have a Canon i9900 13"x19"? Never really tried to print that size as I don't have any images that large at 300dpi

Off to bed.
Thanks

 

[pfogle]

I tend to print (on an Epson 1290) at about 240 pixels per inch, and let the printer driver do the necessary interpolation to the print size. It seems to work fine, I can't really see any difference between that, and resizing to 300 dpi manually.

I've never been asked for a 'dpi' from a client, but I have been asked for a 50MB file (from Getty). Since the R-D1 produces roughly 36MB files max (16 bit tiff from RAW) I'd have to upsize to fulfil the brief.

What I have noticed, is that when printing near the limit for the file size, small hightlights in particular show the pixels, and lines can show artifacts (moire and 'jaggies'), so using the interpolation feature will smooth out those features, and give a better looking print.

ps lightwriter... I expect you'll get superb results from your printer, from what I've read in the reviews.

btw, I'm currently travelling with a Sony notepad and the R-D1. I've bought a Canon ip4200 for doing proofs, but it's not very good for color (way saturated and red). Does anyone have any advice on using this printer or know of an easy way of calibrating?

thanks
Phil

 

[jlw]

One final question though as am I still a bit confused. If one needs 1.4x as many pixels as dots to prevent moire and someone asks for a 300 dpi image - wouldn't you take the 3600x3600 pixel image and divide by 1.4 first and then by 300 which gives an image 8.57 inches square?

It's going to drive people crazy if we hijack this thread into a discussion of printing esoterica, so I'll try to keep this short:

Let's take the case of the 3600x3600-pixel image that somebody wants to print 12 inches square. 3600 / 12 tells me this is printing at a pixel density of 300 pixels per inch at the final printed size (which is what counts in this calculation.)

Now, given that I'm sending 300 pixels per inch to the imagesetter, dividing by 1.4 tells me that I could print using a halftone linescreen of 214 dots per inch (300 / 1.4.) That's a much finer linescreen than a printer would use even for high-quality magazine reproduction, which typically is 133 to 150 dots per inch. So, I'd know my image has plenty of data enough for this purpose, plus a safety factor in case the art director gets cute and decides to crop it a bit.

Now let's compare this with a case where I've got an image that's only 1800x1800 pixels. If I try to print THAT at 12 inches square, I'm only getting a linear density of 150 pixels per inch (1800 / 12.) Applying my 1.4 factor tells me that the finest linescreen I can use with this image would be 107 dots per inch (150/1.4.) That means it would be fine for a newspaper (typically 65 dots per inch) or "commercial" offset on bond paper (seldom more than 100 lines per inch) but I'd be pushing my luck using it at that size for high-quality magazine or catalog printing (often 133 or 150 dots per inch.)

I should also tell you that this 1.4 factor is the bare minimum (actually it's rounded off from the square root of 2, which is 1.414.) Lots of designers prefer to use a factor of at least 1.5, and many use a factor of 2 to be safe. That's where you get the often-heard rule of thumb that an image that prints at 300 pixels per inch is safe to use with a linescreen of up to 150 dots per inch.


One more thing and then I absolutely promise I'll stop: Where does this factor of 1.4 or 1.5 or 2 come from? What's behind it is that you want each printed dot to represent MORE than one image pixel. If your image had 100 pixels per inch and you printed it at 100 dots per inch, the pixel edges and dot edges might line up, and then you'd get noise or moiré patterns. The multiplier number is just a way of assuring that each printed dot always contains data from more than one image pixel, which provides "supersampling" and helps smooth out the printed image.

That's positively all I'll say about this. If anyone wants to debate/discuss it further, we probably should start a new thread... although actually, since it isn't RFF-related, we might not oughtta be discussing it here at all! Sorry I opened this can of worms...