Megapixels, Noise, Image Quality, and Fixing Photos in Software
- You Can Improve Bad Photos in Photoshop
- Sufficient Quality
- Megapixels and Noise
- iMac or Mac mini?
- Cheap USB WiFi and Failing SuperDrives
- Macs in Philly
- Optical Mice for ADB Macs
Enjoyed this article Dan. One quibble though. You said:
"Those shot with an 8 MP compact camera look horrid: low contrast, washed out colors, more grain, and blurring of details. And there's nothing you can do in Photoshop to fix those problems."
Actually you can fix some of those problems in Photoshop. Arguably, you can fix all of them to some extent. Low contrast and washed out colors can be fixed relatively easily in Photoshop while the excessive noise (or grain) and blurriness present more of a problem, but both can be addressed to an extent (although fixing the noise would tend to increase the blurriness and vice-versa). That being said, who wants to spend a lot of time fixing these issues for every photo you shoot?
To be clear, I agree with you completely that the increasing pixel count in point and shoot cameras is a problem and I appreciate your efforts to educate casual users on this issue. I've been quite happy with the 5 megapixel Canon S2 I bought a bit over a year ago, and believe I can get better results with it than I would with the newer models (it's up to S5) that go up to 8 megapixels (I believe).
Enjoy the site and the services you provide to the Mac community.
I have firsthand experience: my sister-in-law got a great new Canon digital last year and offered to shoot our wedding. 8 MP and lots of grain, washed out highlights, noisy shadows. I can make them better in Photoshop, but not close to what they should be. You can fix some things, but increasing contrast boosts noise, decreasing noise reduces sharpness, and in the end you're better off buying a better camera than wasting so much of your valuable time fixing things.
Waverly and I bought a Kodak EasyShare 8 MP SLR-style digital for our honeymoon. The only disappointment, which it shares with almost all non-SLR camera, is horribly limited flash range. (Most low-end film cameras had the same problem, but we could always suggest Fujicolor 800 or 1600 to help out.) Anyhow, except for flash range, the Kodak takes great photos.
I take the "low end" philosophy seriously: Do your research and buy something that's good enough now and for the foreseeable future. Quality lasts; cheap stuff doesn't.
From Eric Matthieu:
Amen to all that you wrote in today's column. Without a doubt, I would still be happily taking snapshots with a 3.1 MP Nikon Coolpix 3100 if it weren't for dropping it last Fall. The sharpness of pictures taken with it is hard to surpass (unless you take into account those taken with any zoom). At four years old, it still did a great job for most of my needs and was very reliable. I rarely print pics, but when I do it's usually at 4x6. Most of the time, my pictures are emailed to others to enjoy.
After canvassing the Internet, I replaced the Nikon with an 8 MP Panasonic (Lumix DMC-TZ3). It's a great little camera, and comparing it to the old Nikon, has a much better LCD, excellent video capture (with sound!), and of course there is its wonderful Leica lens (10x optical zoom; 28-280mm equivalent). I am very pleased with my new toy. With all of the advancements made over the past few years, it's no wonder!
But you hit the nail on the head when it comes to snapshots. Anything more is really overkill for most pictures. At least for amateurs like me.
Hopefully at least some manufacturers will address the septum issue you mentioned. They can't all eliminate it, can they? Maybe there will be a few who either retain it or somehow otherwise compensate for its loss (I'm probably showing my ignorance here, but I think you get my point). People are interested in quality photographs more than specs . . . right?
It's a shame, but most consumers don't do a lot of research before plunking down $500, $300, or even $100 for a digital camera. They want a lot of megapixels, a lot of zoom range, and a big display. They don't care about lens speed or flash range until it's too late. They don't read online reviews and look at photos taken with the camera in less-than-ideal conditions. They believe that what they see on the 3" LCD is indicative of final output quality, when it's really like comparing a postage stamp to a 4x6 print.
I've worked in camera stores on and off since high school, and I love educating my customers. I'd rarely tell them what to buy. Instead I'd tell them why they wanted a fast lens, a zoom lens, a wide lens, a powerful flash, or whatever else would help them take better pictures. Alas, most people don't buy from people who can educate them about their options.
Well, enough for that rant. We're among a small group of photographers doing our own myth busting, and my hat is off to Adorama for not only talking about the "no septum" problem but photographically documenting it.
I think 98% of photographers would be completely happy with 4-8 MP cameras, but they're being lead down the wrong path by marketers. We just need to educate them, explaining and showing them when possible the problems they want to avoid.
From Walt French:
As you say, the proof's in the pudding, and the example you show, comparing a DSLR 8 MP photo with a 8 MP point-n-shooter confirms that it's not megapixels, but rather other factors.
Simple physics dictates that the larger the sensor and the lower the ASA, the better the image. Still, I'll wager that the example you published didn't compare a Canon EOS or Rebel versus a Canon 850IS or similar pocket camera; engineering matters.
I fear that the rant on megapixels distracts from users considering the quality of the photos under their own shooting conditions. In fact, in good light or flash, a tiny point'n'shooter can take exceptional pictures. And the new, multi-megapixel cameras will take better photo than their 3-year-old, half-as-many MP brethren.
In truth (and I say this as someone who's involved with developing software for processing the raw images), more megapixels is better, all else equal. I'd much rather have a chip that's chopped up into 12 MP than one divided into 6 MP. Yes, each pixel will be noisier, but by giving me more information, I (or my software) can decide whether to average 'em back together or recognize that one cell is red and the other is blue. Under many light conditions, I'll make that decision very accurately; under others, I'll have to fall back on the average. Choice is good - in Finance, my other field, every option has value - and you are tilting at windmills by saying more data, even if noisier, is bad.
Why not simply encourage readers to look at the pix and make their own decisions, giving them pointers at what to look at critically?
And BTW: the link about DSLR vs 35mm is hopelessly outdated. As in, "FALSE." Excellent DSLRs from Canon & Nikon are available at $699 list, and the photos from these will often be far superior to 35. For example, both can do a superb job with ASA 1600 equivalence, something utterly impossible with today's best B/W, let alone color films.
Those "hopelessly outdated" articles have dates (from 2003) for a reason - digital is a moving target. It's a topic I need to address again, especially in light of the latest Nikon DSLR. In the realm of "large sensor" cameras (the four-thirds system and bigger), film has effectively been killed.
We're not opposed to 12 MP digicams; we are opposed to digicams that make so many compromises that you're going to end up with muddy or grainy or washed out - whether that's a 1.3 MP cell phone camera or a 12 MP "high end" point-n-shoot.
As far as image quality goes, twice as many pixels with twice the noise is going to look worse no matter how you process it. Noise is random, and you can't make it go away by averaging lots of pixels together. You can reduce its impact, but you can't eliminate it. You can't make a silk purse from a sow's ear - or a crisp image from a soft, noisy one.
Of course, ideally you don't buy the camera with the septum-free imager....
On Feb 26, 2008, at 12:17 PM, Dan Knight wrote:
twice as many pixels with twice the noise is going to look worse no matter how you process it.
Dirty little secret: for a given chip material/engineering, noise is approximately fixed per square mm while signal is proportional to the area. Averaging two cells' noisy signals is identical to combining them back into a single cell. So more megapixels causes no more noise per area.
Except that signal processing is smart about noise these days and doesn't have to just average the two cells. Stay tuned, or look into the elaborate algorithms used to convert RAW files into full color images. The cheapos don't have two DSP locomotives, but they're getting smarter by the day.
And by the way, this septum issue . . . sure, I can see how it matters, but it's merely one of dozens of engineering compromises (or is that redundant?) that any camera manufacturer makes. On the face of it, there are many more important ones, because again, software can compensate to some extent, if you care to, while lens flare, for example, is very difficult to remove.
If there were a single issue I'd advise people on image quality of point'n'shooters, it'd be to have them focus on the area of the sensor . . . that's the primary determinant of quality in low light or high speed snaps. Let them know that a 24mm2 sensor is going to take noisier, less appealing photos than a 36mm2 sensor, regardless of pixel count. For reference, 35mm is 864mm2, 36 times as big as the ultra compacts.
And the fact is, you can often get pretty darn good shots out of a camera that has 1/36 the size of a full frame and about as many megapixels, more than good enough for gummint work. Pretty amazing, no? And where's the megapixel myth if the point'n'shooter doesn't scale the full-frame down to only 1/3 megapixel, so each sensor cell is the same as full frame?
Agreed: Signal is proportional to area while noise is constant. Assuming a sensor capable of 10 bits per pixel and 2 bits of noise at 3 MP, that still gives us 8 bits of meaningful information per pixel. Go to 6 MB, which halves the area and costs us one bit of signal, and you drop to 7 good bits and 1 bit of noise. Now go to 12 MB, which costs us one more bit of signal, and we end up with 6 good bits per pixel and 2 bits of noise.
You're telling me that by averaging together two pixels with 6 good bits and 2 random ones (the noise), you're going to have the same quality as from a larger pixel with 7 good bits and 1 random one. I don't buy it. Averaging two noisier signals is not the same as having one cleaner signal.
And, as you point out, that's the benefit of a bigger sensor. Larger pixels means more signal, which minimizes the effect of noise. Every time you double the number of pixels without changing imager size, you double the impact of noise on your images.
It's a shame manufacturers don't give us meaningful information about sensor size - 1/1.8 of an inch is a hopelessly garbled fraction that would make my fifth grade math teacher shudder (using decimals in fractions?!?!?), but telling me it's a 14mm or 0.55" sensor is something I can relate to.
From Laurence Vittes:
My wife wants to upgrade her eMac (1 GHz, 1 MG SDRAM, Combo drive). She does web design and graphics work (Flash but no video, although it's possible at some point in the future) and typically runs the big Adobe apps concurrently plus Word, Mail, a browser or two, etc. (no gaming, though)
I admire and have come to rely on your coverage of all the possible options, and the value equations, etc., and think her best bet is either a refurbished Mac mini or iMac - but I can't quite find the right article to really nail it for us. Also, as you don't cover displays, I'm uncertain as how to weigh the value of an iMac display vs. a third party display we'd buy to go with a Mini (along with the poky performance of the eMac, her other major complaint is the size).
Any suggestions would be much appreciated!
We don't cover displays, because Apple's have traditionally been very overpriced. Besides, in this age of digital video, it's not really going to matter whether the reviewer is using a Mac or PC. The whole field changes so rapidly that last year's best buy may be long gone from the shelves.
Looking at all the factors - CPU speed, SuperDrive, graphics, etc., we're looking at $679 + a display for a refurbished Mac mini at 2.0 GHz with a pedestrian laptop hard drive and Intel graphics vs. $1,049-1,099 for a 20" 2.0-2.16 GHz iMac with dedicated graphics, a 3.5" hard drive, a built-in iSight webcam, and a new mouse and keyboard. I think the iMac is worth the roughly $400 difference in price. (If you already had a monitor you really liked, the mini would probably win.)
If your wife does graphics, the first question is whether she wants a matte or glossy display, which translates to a 2006 or 2007 iMac. The aluminum iMacs have rich, vibrant color, but they may not be as accurate - and then there's the issue of glare from a flat, glossy surface. I can't make that call, but if you have a Mac dealer nearby, compare the new aluminum iMacs with Apple's Cinema Displays, which have a matte finish like last year's iMacs.
iMac vs. Mac mini isn't something we've really addressed, as they have different markets. The iMac is for someone who wants an all-in-one solution and better performance; the mini is for someone who has a mouse, monitor, and keyboard they want to hold on to or wants a very small computer.
From Nick Byrne:
Longtime reader here. I also love the site, and have picked up a number of tips over the years which have helped me resurrect a variety of aging (and some relatively modern) machines. Two points came up in this week's mailbag to which I feel I may be able to contribute:
The first is that Realtek now has the version 1.5.1 of their drivers for the RTL 8187B chipset up on their website. I also bought a cheap-o TrendNet TEW-424UB adaptor a couple of months ago ($1.99 at CompUSA!), but was unable to get it to work until now. I tried the drivers out on my MacBook (bought for $310 with a cracked screen, so still legitimately Low End™), and it works fine under 10.5.2. If you can get past the terrible UI, utter lack of documentation, and occasional charming misspelling. Frankly, I'm just glad they pay any attention to us Mac users at all.
Another relatively cheap Mac-compatible USB adaptor is the Ativa AWGUA54 using the ZyDAS ZD1211B chipset, which has Mac drivers floating around the Web. Only worth it if you can find it on sale or used.
The second is on the question of failing SuperDrives. The more adventurous with this problem can try removing the drive, looking for missing case screws along the way. Once the drive is out, you can remove the top cover and look for any small screws (or other pieces of metal) which have stuck to the magnetic sled carrying the optical sensor (or elsewhere near the spindle). I have had this happen on at least two Combo drives, both on 12" PowerBooks. I don't know if this could be the same problem, but the description of the sounds appears similar.
Anyway, no need to quote this long email, but I hope this may help some people.
Thanks for the information. I'll post this in the Mailbag in hopes others will benefit from your discoveries.
From Chris Flood:
While it's not an official Apple Store, Springboard Media's shop at 2212 Walnut St. sells nothing but Apple merchandise. Anyone should be able to walk in there and get a Mac Pro, MacBook, iPod, or any other Apple item they want.
Thanks for the reminder. As wonderful as Apple retail stores are, the company would not have survived the mid-to-late 90s without hundreds of local retailers keeping the faith.
From Sebastian Blanco:
I have seen your article about ADB optical mice and the impossibility to get one.
I was thinking about the problem some months ago. I have a lot of old Macs, and I love to use optical mice with them in place of the clunky ball ones.
Well, then I have bought the geethree.com ADB/PS2 adapter. In the website it states that it doesn't work with optical mice.
Like you say in your article, it seems that the problem is there's not enough power to feed the mice optical circuitry in the ADB port.
But optical tech has progressed a lot, and the circuits now draw a lot less energy.
I work in I.T. departments of large companies, and a lot of hardware pass thru my hands, so I was able to test 17 models of PS2 optical mice. I have fund one model that works perfectly using the geethree adapter :D.
Probably it draw little power from the PS2 port so the it works. It is M800-P2M BenQ mouse.
Thanks for the research you did and sharing your findings. The BenQ mouse seems to be widely available around the world - but not in North America.
I haven't been able to find a multi-button optical mouse for use with ADB Macs, but I did buy a Kensington Mouse-in-a-Box USB/ADB, a single-button optical mouse with a USB-to-ADB adapter that only works with this mouse. It's a very lightweight mouse, and I have to admit that until today I hadn't even tried it with the adapter. (I rarely touch ADB Macs these days.) I plugged it into my Blue & White Power Mac, which didn't recognize it in Mac OS X 10.4 until I restarted - but now it does.
I have been using this mouse as my backup mouse for when my wireless Logitech mouse needs to sit in its charging station. It's nice enough, although I do miss having a scroll wheel when I use it. I did a Google product search and found it available from Power Max for $19.88 and Amazon.com for $19.08 (buy two or add a $6 item for free shipping).
Dan Knight has been publishing Low End Mac since April 1997. Mailbag columns come from email responses to his Mac Musings, Mac Daniel, Online Tech Journal, and other columns on the site.
- Mac of the Day: Original iBook G3, introduced 1999.07.21. Innovative, rugged, heavy, clamshell laptop introduced AirPort and was a huge hit.
- Support Low End Mac
Cult of Mac
Shrine of Apple
The Mac Observer
Accelerate Your Mac
The Vintage Mac Museum
Mac Driver Museum
System 6 Heaven
System 7 Today
the pickle's Low-End Mac FAQ