 |
 | Home | Contact Us | FAQ | Search & Site Map | Link to Us |
 | Sign In | Join | Other 45 Sites in Network |
| PhotoKB HomeDiscussion Groups Digital Photography Digital PhotoDSLR CamerasZLR CamerasPoint & Shoot CamerasFilm Photography 35 mmLarge FormatMedium formatDarkroomFilm and LabsOther EquipmentPhoto Technique Nature PhotographyPeople PhotographyTechnique GeneralGeneral Photo Topics General TopicsAustralian PhotographyUK PhotographyDirectoryPhoto Clubs |
Photo Forum / Digital Photography / DSLR Cameras / February 2010On my Nikon D90, what's the difference between fine, normal, and basic image quality?
I can select an image size of large (12.2 megapixel (MP)), Medium (6.9 MP), or Small (3.1 MP). That I understand. I can also select an image quality of fine, normal, or basic. What does this mean? If I set the image size to "large", I find that a typical "fine" image takes about 6.5 megabytes (MB), a "normal" image takes about 4 MB and a "basic" image takes less than 3 MB. What am I losing by selecting normal or basic? > I can select an image size of large (12.2 megapixel (MP)), Medium > (6.9 MP), or Small (3.1 MP). That I understand. [quoted text clipped - 4 lines] > takes about 4 MB and a "basic" image takes less than 3 MB. What am I > losing by selecting normal or basic? You are losing quality, as you're not capturing the maximum number of pixels, and lossy compression degrades any image. Mind you, none of this will be apparent at small sizes. The general wisdom is to capture the full enchilada, and downsize a copy for web, mail, etc.  Signaturejohn mcwilliams > I can select an image size of large (12.2 megapixel (MP)), Medium > (6.9 MP), or Small (3.1 MP). That I understand. [quoted text clipped - 4 lines] > takes about 4 MB and a "basic" image takes less than 3 MB. What am I > losing by selecting normal or basic? Quality is the level of JPEG compression. JPEG is a "lossy" file, meaning that for each level of compression, which will give you a smaller file size there is a loss of image quality which is not recoverable. With your D90 you have the option of recording the image as a JPEG (with the 3 levels of compression, or quality), RAW (Nikon NEF, the lossless highest quality file, your so called "Digital negative"), RAW + JPEG, or if IRC TIFF (another lossless file type) With the current low prices of 8GB or 16GB SD cards I would recommend shooting in RAW or RAW + JPEG(fine) so that you have the the best quality image file for processing on your computer. That way you can size, or crop for whatever purpose you choose and maintain image quality. SignatureRegards, Savageduck >I can select an image size of large (12.2 megapixel (MP)), Medium >(6.9 MP), or Small (3.1 MP). That I understand. > >I can also select an image quality of fine, normal, or basic. What >does this mean? It will set the JPEG compression rate. JPEG compression is not lossless and the more you compress a JPEG the more artifacts are introduced resp. the more details are lost. jue > I can also select an image quality of fine, normal, or basic. What > does this mean? JPEG compresses lossily. You choose the tradeoff between file sizes and irrecoverable loss of certain details and addition of artifacts. Try it out, especially with fine detail (e.g. details in fine hair). JPEG is designed to throw away first invisible and near invisible details, so you usually need to look closely and at 100% or more. It's usually better to lower the quality than reducing the number of pixels --- pixels aren't really 'near invisible details', and unless you *know* you'll never need the image printed/displayed larger than the pixels can deliver, ... On the other hand I remember someone testing that you loose practically no data (as a fourier transformation shows) when downsizing a Bayer-pattern image to 70% (i.e. 70% as high and 70% as wide) due to the way Bayer pattern sensors record detail. -Wolfgang >> I can also select an image quality of fine, normal, or basic. What >> does this mean? > JPEG compresses lossily. You choose the tradeoff between file > sizes and irrecoverable loss of certain details and addition > of artifacts. > Try it out, especially with fine detail (e.g. details in fine > hair). JPEG is designed to throw away first invisible and near > invisible details, so you usually need to look closely and at 100% > or more. > It's usually better to lower the quality than reducing the number > of pixels --- pixels aren't really 'near invisible details', and > unless you *know* you'll never need the image printed/displayed > larger than the pixels can deliver, ... > On the other hand I remember someone testing that you loose > practically no data (as a fourier transformation shows) when > downsizing a Bayer-pattern image to 70% (i.e. 70% as high and > 70% as wide) due to the way Bayer pattern sensors record detail. I find that's true with most of my zoom lens photographs, but a really good prime used with care to get maximum detail usually can't be downsized without losing detail. In fact I use experimental downsizing by 33% as my test of whether I've managed to exploit the lens resolution to the max. SignatureChris Malcolm >> On the other hand I remember someone testing that you loose >> practically no data (as a fourier transformation shows) when >> downsizing a Bayer-pattern image to 70% (i.e. 70% as high and >> 70% as wide) due to the way Bayer pattern sensors record detail. > I find that's true with most of my zoom lens photographs, but a really > good prime used with care to get maximum detail usually can't be > downsized without losing detail. In fact I use experimental downsizing > by 33% as my test of whether I've managed to exploit the lens > resolution to the max. So you have fine detail to frequencies of 1 pixel per half wave in your shots? -Wolfgang >>> On the other hand I remember someone testing that you loose >>> practically no data (as a fourier transformation shows) when >>> downsizing a Bayer-pattern image to 70% (i.e. 70% as high and >>> 70% as wide) due to the way Bayer pattern sensors record detail. >> I find that's true with most of my zoom lens photographs, but a really >> good prime used with care to get maximum detail usually can't be >> downsized without losing detail. In fact I use experimental downsizing >> by 33% as my test of whether I've managed to exploit the lens >> resolution to the max. > So you have fine detail to frequencies of 1 pixel per half > wave in your shots? I don't need to know the answer to that question. It's also possible that it contains some theoretical simplifying assumptions which oversimplify the issue of visible detail in photographic images. All I need to know is that if I downsize by 33% and then compare the result with the original there is obvious loss of detail, such as fine twigs blurring, distant street names and car numberplates becoming harder to read, distant roof tiling becoming more obscure, etc.. SignatureChris Malcolm >> So you have fine detail to frequencies of 1 pixel per half >> wave in your shots? > All I > need to know is that if I downsize by 33% and then compare the result > with the original there is obvious loss of detail, such as fine > twigs blurring, distant street names and car numberplates becoming > harder to read, distant roof tiling becoming more obscure, etc.. Take any true-type (i.e. vector) font and look at it 33% smaller (if only by increasing the observation distance). Of course it'll be harder to read, even though no loss of detail has happened. I also find myself often enlarging images (e.g. from other people) to 200% or 400% to better judge fine detail --- even though no more detail is being made visible. Hence your test, at least as I understand it as written here, is not usable to judge loss of detail versus worse presentation of detail. Thus my question, what high frequencies are encoded in your pixels that downsampling will hurt them. (Observe that your camera probably has an anti-aliassing filter, which also blurs and attenuates very high frequency details --- as it must to avoid aliassing ...) -Wolfgang >>> So you have fine detail to frequencies of 1 pixel per half >>> wave in your shots? >> All I >> need to know is that if I downsize by 33% and then compare the result >> with the original there is obvious loss of detail, such as fine >> twigs blurring, distant street names and car numberplates becoming >> harder to read, distant roof tiling becoming more obscure, etc.. > Take any true-type (i.e. vector) font and look at it 33% smaller > (if only by increasing the observation distance). Of course it'll > be harder to read, even though no loss of detail has happened. I avoid that effect by comparing them at the same viewing size, using simple pixel magnification. > I also find myself often enlarging images (e.g. from other people) > to 200% or 400% to better judge fine detail --- even though no > more detail is being made visible. Exactly. In order to be sure of what is going on I will often make the pixels visible. SignatureChris Malcolm >> Take any true-type (i.e. vector) font and look at it 33% smaller >> (if only by increasing the observation distance). Of course it'll >> be harder to read, even though no loss of detail has happened. > I avoid that effect by comparing them at the same viewing size, using > simple pixel magnification. "simple" magnification (e.g. nearest neighbour) will give suboptimal results. >> I also find myself often enlarging images (e.g. from other people) >> to 200% or 400% to better judge fine detail --- even though no >> more detail is being made visible. > Exactly. In order to be sure of what is going on I will often make the > pixels visible. Ans that can be self-defeating again. In most of my photographs I'm not concerned about visible pixels. Instead, I'd be rather concerned if pixels were visible. :-) Only with that understanding can one undertake the trip to 100+% views and come out wiser --- and if some frame does not contain high frequency details, downsampling it so it only encodes the frequencies it contains does not cause appreciable data loss. (Remember that while interpolating from bayer sensors does give very good results, it cannot encode data it doesn't see without inventing it. Which is one reason why RAW (outside Foveon) doesn't save RGB triplets.). -Wolfgang >>> Take any true-type (i.e. vector) font and look at it 33% smaller >>> (if only by increasing the observation distance). Of course it'll >>> be harder to read, even though no loss of detail has happened. >> I avoid that effect by comparing them at the same viewing size, using >> simple pixel magnification. > "simple" magnification (e.g. nearest neighbour) will give > suboptimal results. That can be true in certain cases at certain magnifications. Easy enough to check that and sidestep any misleading artefacts by simply shifting up and down the scale. >>> I also find myself often enlarging images (e.g. from other people) >>> to 200% or 400% to better judge fine detail --- even though no >>> more detail is being made visible. >> Exactly. In order to be sure of what is going on I will often make the >> pixels visible. > Ans that can be self-defeating again. In most of my photographs > I'm not concerned about visible pixels. > Instead, I'd be rather concerned if pixels were visible. :-) Of course. But to check out pixel level artefacts and find out where they're coming from you have to look at the pixels. Just as although you have no aesthetic interest photographs of test charts, they can come in handy when checking out a lens. > Only with that understanding can one undertake the trip to 100+% > views and come out wiser --- and if some frame does not contain > high frequency details, downsampling it so it only encodes the > frequencies it contains does not cause appreciable data loss. Exactly. That's why I use downsampling as a quick test of the resolution achieved in a particular image, and why I use simple downsampling amounts which correspond to simple vulgar fractions. > (Remember that while interpolating from bayer sensors does give > very good results, it cannot encode data it doesn't see without > inventing it. Which is one reason why RAW (outside Foveon) > doesn't save RGB triplets.). Of course. And one reason for not saving RGB triplets from Bayer arrays is that there is more detail resolution than colour resolution in the array which any specific triplet translation method necessarily compromises. SignatureChris Malcolm >>> I avoid that effect by comparing them at the same viewing size, using >>> simple pixel magnification. >> "simple" magnification (e.g. nearest neighbour) will give >> suboptimal results. > That can be true in certain cases at certain magnifications. Even in all magnifications in certain cases. > Easy > enough to check that and sidestep any misleading artefacts by simply > shifting up and down the scale. Better to use a competent magnification in the first place. And interesting to use some magnification method that invents details, just to countercheck the assumption that you see true detail in all cases. >>> Exactly. In order to be sure of what is going on I will often make the >>> pixels visible. >> Ans that can be self-defeating again. In most of my photographs >> I'm not concerned about visible pixels. >> Instead, I'd be rather concerned if pixels were visible. :-) > Of course. But to check out pixel level artefacts and find out where > they're coming from you have to look at the pixels. True. > Just as although > you have no aesthetic interest photographs of test charts, they can > come in handy when checking out a lens. Also true. Though I seem to remember at least some test charts coming from "how much detail can one resolve in aerial photography (i.e. what can one detect) (and what is just the effect of film emulsion and humans 'seeing' things)" rather than aesthetic interests. ;-) In other words, pixel peeping is fine ... within limits. > That's why I use downsampling as a quick test of the > resolution achieved in a particular image, and why I use simple > downsampling amounts which correspond to simple vulgar fractions. I'd rather find out what frequencies are in the image (fourier transform) rather than downsampling (which has to be done competently, again). But that's me. >> (Remember that while interpolating from bayer sensors does give >> very good results, it cannot encode data it doesn't see without >> inventing it. Which is one reason why RAW (outside Foveon) >> doesn't save RGB triplets.). > Of course. And one reason for not saving RGB triplets from Bayer > arrays is that there is more detail resolution than colour resolution > in the array "Green" (which is actually not too insensitive for red and blue) carries most resolution, and occupies 50% of the array. It's distance to the next green pixel is SQRT(2), which is ca. 1/70% ... > which any specific triplet translation method necessarily > compromises. Of course you are wrong here. Just generate a TIFF from your RAW with the same numbers of pixels, just store RGB triplets instead of place-dependent luminosity-behind-$COLORFILTER-values. Presto, a triplet translation which doesn't loose data (it's justr wasteful) and can be turned into JPEGs, too. -Wolfgang >>>> I avoid that effect by comparing them at the same viewing size, using >>>> simple pixel magnification. >>> "simple" magnification (e.g. nearest neighbour) will give >>> suboptimal results. >> That can be true in certain cases at certain magnifications. > Even in all magnifications in certain cases. But not in cases where the magnification is simply of only a single pixel, such as 4 times magnification, where one pixel becomes represented by four of the same value. And so on. That's why I choose that kind of very simple magnification for viewing magnification changes. >> Easy >> enough to check that and sidestep any misleading artefacts by simply >> shifting up and down the scale. > Better to use a competent magnification in the first place. Which is as I explained what I do. It never interpolates or blends when the magnification factor does not demand it. > And interesting to use some magnification method that invents > details, just to countercheck the assumption that you see true > detail in all cases. I wouldn't find that interesting, which is why I don't do it. >>>> Exactly. In order to be sure of what is going on I will often make the >>>> pixels visible. >>> Ans that can be self-defeating again. In most of my photographs >> That's why I use downsampling as a quick test of the >> resolution achieved in a particular image, and why I use simple >> downsampling amounts which correspond to simple vulgar fractions. > I'd rather find out what frequencies are in the image > (fourier transform) rather than downsampling (which has to be > done competently, again). But that's me. Fourier transforms don't work for me because they can't tell the difference between image detail and noise. >>> (Remember that while interpolating from bayer sensors does give >>> very good results, it cannot encode data it doesn't see without >>> inventing it. Which is one reason why RAW (outside Foveon) >>> doesn't save RGB triplets.). >> Of course. And one reason for not saving RGB triplets from Bayer >> arrays is that there is more detail resolution than colour resolution >> in the array > "Green" (which is actually not too insensitive for red and > blue) carries most resolution, and occupies 50% of the array. > It's distance to the next green pixel is SQRT(2), which is > ca. 1/70% ... I was talking about non-chromatic detail, which doesn't have to reach as far as the next sensor of the same colour. >> which any specific triplet translation method necessarily >> compromises. > Of course you are wrong here. Just generate a TIFF from your RAW > with the same numbers of pixels, just store RGB triplets instead > of place-dependent luminosity-behind-$COLORFILTER-values. > Presto, a triplet translation which doesn't loose data (it's > justr wasteful) and can be turned into JPEGs, too. You might be right there. I've never played with TIFFs. I have noted that different RAW converters give different levels of detail resolution, in some special cases quite marked. And sometimes a RAW converter will offer an update with improved translation for the RAW from a specific camera, and the reviewers exclaim at the extra detail it gets. SignatureChris Malcolm >>>> "simple" magnification (e.g. nearest neighbour) will give >>>> suboptimal results. >>> That can be true in certain cases at certain magnifications. >> Even in all magnifications in certain cases. > But not in cases where the magnification is simply of only a single > pixel, such as 4 times magnification, where one pixel becomes > represented by four of the same value. Nearest neighbour --- and that's what it is --- will give suboptimal results for most viewing cases. This includes upsampling for compatison with images with more pixels. > And so on. That's why I choose > that kind of very simple magnification for viewing magnification > changes. Take a smooth gradient between black and white. Downsample it, upsample (or magnify) your way and you get banding. Upsample properly and you get no banding. >>> Easy >>> enough to check that and sidestep any misleading artefacts by simply >>> shifting up and down the scale. >> Better to use a competent magnification in the first place. > Which is as I explained what I do. It never interpolates or blends > when the magnification factor does not demand it. I respectfully disagree --- your choosen magnification is incompetent for photography and twice incompetent for comparing photographs of different pixel density. >> And interesting to use some magnification method that invents >> details, just to countercheck the assumption that you see true >> detail in all cases. > I wouldn't find that interesting, which is why I don't do it. Of course you do, but the details your magnification invents (banding, borders, hard change) are rather drab >> I'd rather find out what frequencies are in the image >> (fourier transform) rather than downsampling (which has to be >> done competently, again). But that's me. > Fourier transforms don't work for me because they can't tell the > difference between image detail and noise. That must be strong noise if it overpowers detail even in test shots designed for maximum detail and minimum noise. >> "Green" (which is actually not too insensitive for red and >> blue) carries most resolution, and occupies 50% of the array. >> It's distance to the next green pixel is SQRT(2), which is >> ca. 1/70% ... > I was talking about non-chromatic detail, which doesn't have to reach > as far as the next sensor of the same colour. And what happens to that detail once it hits the AA filter? > I have noted > that different RAW converters give different levels of detail > resolution, in some special cases quite marked. It depends a lot on how conservative they are with the input --- if they don't mind guessing, they are going o have more detail (and more *false* detail) in their results. Inventing details can be a workable strategz for: > And sometimes a RAW > converter will offer an update with improved translation for the RAW > from a specific camera, and the reviewers exclaim at the extra detail > it gets. such accolades. -Wolfgang >>>>> "simple" magnification (e.g. nearest neighbour) will give >>>>> suboptimal results. >>>> That can be true in certain cases at certain magnifications. >>> Even in all magnifications in certain cases. >> But not in cases where the magnification is simply of only a single >> pixel, such as 4 times magnification, where one pixel becomes >> represented by four of the same value. > Nearest neighbour --- and that's what it is --- will give > suboptimal results for most viewing cases. This includes > upsampling for compatison with images with more pixels. >> And so on. That's why I choose >> that kind of very simple magnification for viewing magnification >> changes. > Take a smooth gradient between black and white. Downsample it, > upsample (or magnify) your way and you get banding. Upsample > properly and you get no banding. >>>> Easy >>>> enough to check that and sidestep any misleading artefacts by simply >>>> shifting up and down the scale. >>> Better to use a competent magnification in the first place. >> Which is as I explained what I do. It never interpolates or blends >> when the magnification factor does not demand it. > I respectfully disagree --- your choosen magnification is > incompetent for photography and twice incompetent for comparing > photographs of different pixel density. >>> And interesting to use some magnification method that invents >>> details, just to countercheck the assumption that you see true >>> detail in all cases. >> I wouldn't find that interesting, which is why I don't do it. > Of course you do, but the details your magnification invents > (banding, borders, hard change) are rather drab I think there's a confusion going on between upsampling and downsampling for purposes of viewing an image, and doing so in order to change the size of the image in pixels. As I explained earlier I use different methods for those two different purposes. >>> I'd rather find out what frequencies are in the image >>> (fourier transform) rather than downsampling (which has to be >>> done competently, again). But that's me. >> Fourier transforms don't work for me because they can't tell the >> difference between image detail and noise. > That must be strong noise if it overpowers detail even in test > shots designed for maximum detail and minimum noise. I thought you were more interested in real images than pixel peeping on test charts taken in ideal conditions? In the real images I deal with compromise between reducing noise and preserving detail resolution is an everyday problem. >>> "Green" (which is actually not too insensitive for red and >>> blue) carries most resolution, and occupies 50% of the array. >>> It's distance to the next green pixel is SQRT(2), which is >>> ca. 1/70% ... >> I was talking about non-chromatic detail, which doesn't have to reach >> as far as the next sensor of the same colour. > And what happens to that detail once it hits the AA filter? You're asking theoretical questions and giving answers based on simplifying mathematical assumptions when I presume you actually do have a good DSLR and at least one lens with sufficiently good central precision that critical cases can be found where downsizing the image by 30% loses some central detail resolution? My impression is that that there's a few such lenses in the lens armoury of most DSLR makes. I'd be prepared to make an exception if your camera is the Canon EOS 7D :-) SignatureChris Malcolm >>>> And interesting to use some magnification method that invents >>>> details, just to countercheck the assumption that you see true >>>> detail in all cases. >>> I wouldn't find that interesting, which is why I don't do it. >> Of course you do, but the details your magnification invents >> (banding, borders, hard change) are rather drab > I think there's a confusion going on between upsampling and > downsampling for purposes of viewing an image, and doing so in order > to change the size of the image in pixels. Explain to me the difference between up/downsampling an image to view or print it at a certain number of pixels and up/downsampling an image to a certain number of pixels (with the assumption that the image *will* be viewed or printed at some time, maybe with another up/downsampling step, as needed). > As I explained earlier I > use different methods for those two different purposes. I don't see the difference, sorry >>> Fourier transforms don't work for me because they can't tell the >>> difference between image detail and noise. >> That must be strong noise if it overpowers detail even in test >> shots designed for maximum detail and minimum noise. > I thought you were more interested in real images than pixel peeping > on test charts taken in ideal conditions? I am. > In the real images I deal > with compromise between reducing noise and preserving detail > resolution is an everyday problem. If the very best test shots in ideal circumstances only resolve X, how do you expect to resolve more than X in the ordinary case? And how can you then not recognize that if you only resolve X, anything with higher frequencies must be noise, and hence can be dropped without regret? >>> I was talking about non-chromatic detail, which doesn't have to reach >>> as far as the next sensor of the same colour. >> And what happens to that detail once it hits the AA filter? > You're asking theoretical questions If you insist ... "Please provide a test shot in ideal conditions with *your* gear, made to contain the highest frequencies you can record with your camera, as noise free as you can make them. Then run a FFT over that and kindly tell us what the highest frequencies are that you can resolve. Thank you." And kindly advise how that non-chromatic detail, which is spread by a competent AA filter, doesn't reach as far as the next sensor of the same colour ... and if it doesn't reach the next sensor, how can it be displayed in correct colour? Assume that for all relevant cases the green pixels can detect the non-chromatic detail, and that their distance is 1.4 pixels from each other. -Wolfgang >>>>> And interesting to use some magnification method that invents >>>>> details, just to countercheck the assumption that you see true >>>>> detail in all cases. >>>> I wouldn't find that interesting, which is why I don't do it. >>> Of course you do, but the details your magnification invents >>> (banding, borders, hard change) are rather drab >> I think there's a confusion going on between upsampling and >> downsampling for purposes of viewing an image, and doing so in order >> to change the size of the image in pixels. > Explain to me the difference between up/downsampling an image to > view or print it at a certain number of pixels and up/downsampling > an image to a certain number of pixels (with the assumption that > the image *will* be viewed or printed at some time, maybe with > another up/downsampling step, as needed). This is where I came in, I think. I explained that a long time ago. You've confused yourself by your use of the word "number of pixels" in the above paragraph. You're making assumptions I don't make. If by any chance you're really interested in my answer to the above question, go back and read my original explanation. SignatureChris Malcolm >>> I think there's a confusion going on between upsampling and >>> downsampling for purposes of viewing an image, and doing so in order >>> to change the size of the image in pixels. >> Explain to me the difference between up/downsampling an image to >> view or print it at a certain number of pixels and up/downsampling >> an image to a certain number of pixels (with the assumption that >> the image *will* be viewed or printed at some time, maybe with >> another up/downsampling step, as needed). > This is where I came in, I think. Aeh, nope. You came in on fine details. > I explained that a long time > ago. You explained that you prefer nearest neighbour for, ah, better judging details by producing blocky upsizing artifacts. You'll note that that is neither "for viewing an image" nor "to change the size of the image in pixels". > You've confused yourself by your use of the word "number of > pixels" in the above paragraph. So you change the size of an image in pixels, but don't change the numbers of pixels?? Did *you*, by chance, mean 'change the apparent pixel size of an image' (which of course increases the number of pixels, usually by a power of 4)? If so, you confused me by your completely muddled word choice, then pointed to me as the culprit. Good job, really! > You're making assumptions I don't > make. One of my assumptions was that you'd choose your words carefully and intended to express what they convey. Should I not use that assumption when reading what you write? > If by any chance you're really interested in my answer to the > above question, go back and read my original explanation. If by any chance you're really interested in my answer to your original explanation, go read my answer to your original explanation. I understood you perfectly well, thank you very much, but you seem intent on worst case upsampling for pixel peeping. Of course that brings suboptimal results, like presumed resolution loss, if you upsize by e.g. 141% ... -Wolfgang > Try it out, especially with fine detail (e.g. details in fine > hair). JPEG is designed to throw away first invisible and near > invisible details, so you usually need to look closely and at 100% > or more. Try a test with any program that re-saves jpegs at the worst jpeg setting and you'll see the rectangular blocky artifacts in smooth transition areas like skies and rounded human features. The basic setting is really not that bad till you zoom way in or try to lighten the exposure, then you see blocky banding in skies where it simplified the file to save space. I use it if I'm just documenting a label or sign, etc or for time lapse movies where it will get compressed to death anyways and the file sizes get out of control with thousands of images. SignaturePaul Furman www.edgehill.net www.baynatives.com all google groups messages filtered due to spam >> Try it out, especially with fine detail (e.g. details in fine >> hair). JPEG is designed to throw away first invisible and near >> invisible details, so you usually need to look closely and at 100% >> or more. > Try a test with any program that re-saves jpegs at the worst jpeg > setting and you'll see the rectangular blocky artifacts in smooth > transition areas like skies and rounded human features. Which is pretty irrelevant, as the camera doesn't have such a setting (not even near) in first place --- much as a chain saw as "how bad can cutting be" isn't not too relevant to your choice of kitchen knifes (one hopes). > The basic > setting is really not that bad till you zoom way in or try to lighten > the exposure, then you see blocky banding in skies where it simplified > the file to save space. Beware that Nikon can either compress to size or to quality --- the first gives consistent file sizes, but overdoes easy things, but does a bad job on complex images. The second gives changing file sizes instead. > I use it if I'm just documenting a label or > sign, etc or for time lapse movies where it will get compressed to death > anyways and the file sizes get out of control with thousands of images. The OP should really test things out (and then choose one setting better than he thinks is OK for him). Or shoot RAW anyway. -Wolfgang >>> Try it out, especially with fine detail (e.g. details in fine >>> hair). JPEG is designed to throw away first invisible and near [quoted text clipped - 9 lines] > "how bad can cutting be" isn't not too relevant to your choice > of kitchen knifes (one hopes). It is instructional to see the kind of damage compression does on an image by looking at an exaggerated example, otherwise it's quite subtle & hard to detect much less understand. >> The basic >> setting is really not that bad till you zoom way in or try to lighten [quoted text clipped - 14 lines] > > -Wolfgang SignaturePaul Furman www.edgehill.net www.baynatives.com all google groups messages filtered due to spam >I can select an image size of large (12.2 megapixel (MP)), Medium >(6.9 MP), or Small (3.1 MP). That I understand. [quoted text clipped - 4 lines] >takes about 4 MB and a "basic" image takes less than 3 MB. What am I >losing by selecting normal or basic? Other have provided the fish. I prefer to teach you how to fish and feed yourself. RTFM! This and the answer to many other questions you may have may be found in the Fine Manual. http://support.nikontech.com/app/answers/detail/a_id/16087/session/L3NpZC9Zajc3O FdSag%3D%3D/p/19%2C544/c/187/r_id/116678/sno/1 >> I can select an image size of large (12.2 megapixel (MP)), Medium >> (6.9 MP), or Small (3.1 MP). That I understand. [quoted text clipped - 10 lines] > > http://support.nikontech.com/app/answers/detail/a_id/16087/session/L3NpZC9Zajc3O FdSag%3D%3D/p/19%2C544/c/187/r_id/116678/sno/1 What was I thinking? I made the assumption the OP had actually read TFM. You have provided the best advice given in this thread,as all the answers are there. SignatureRegards, Savageduck > I can select an image size of large (12.2 megapixel (MP)), Medium > (6.9 MP), or Small (3.1 MP). That I understand. [quoted text clipped - 4 lines] > takes about 4 MB and a "basic" image takes less than 3 MB. What am I > losing by selecting normal or basic? To keep it simple - yes, the other posts are correct. Shoot the RAW file most of the time, if not ALL of the time with the accompanied "jpg". Use the jpg to mess around with and post / retouch / correct. The RAW file is like having a negative in your drawer - so make sure you back up the keep the RAW files safe on an external or on a CD / DVD. Jeff Roush photo instructor http://www.RoushPhotoOnline.com |
Reply to this Message |
 Sign In
 Join
 My Latest Posts My Monitored Threads My Blog My Photo Gallery My Profile My Homepage Start New Thread Enable EMail Alerts Rate this Thread
|
©2010 Advenet LLC Privacy Policy - Terms of Use
This website includes both content owned or controlled by Advenet as well as content owned or controlled by third parties.