jpg or tiff

[Wingnut]

Just got my first digital camera, FZ3. I'm struggling between using jpeg or tiff. I want to always have the option of 8x10 enlargements & or Photoshop adjustments. Although I'll only need that 4-5% of the time.

I started to shoot everything in Tiff mode but my Pentium 3 takes forever to save & use them. I did download a jpg that a FZ3 took (from this site) & I printed an 8x10. It looked pretty good.

If I stick to JPG fine, will the quality be good enough? Will it break down quickly in PS?

TIA,
Wingnut

[b.j.c]

Just got my first digital camera, FZ3. I'm struggling between using jpeg or tiff. I want to always have the option of 8x10 enlargements & or Photoshop adjustments. Although I'll only need that 4-5% of the time.

I started to shoot everything in Tiff mode but my Pentium 3 takes forever to save & use them. I did download a jpg that a FZ3 took (from this site) & I printed an 8x10. It looked pretty good.

If I stick to JPG fine, will the quality be good enough? Will it break down quickly in PS?

TIA,
Wingnut

When I first got my FZ20 I did some tests with the relative quality of JPG and TIFF files. The TIFF files are huge (17MB) and slow the computer down like you said, and I didn't think the quality was much better. Maybe not better at all. I think the thing with editting JPGs is that everytime you save it, you lose a bit more quality. Thats what I found, anyway. If you convert your file to a TIFF before starting on editting it, then you don't lose quality with every save.

I'm just sticking to high quality JPG.

[genece]

If you need to ask you don't need tiff, I don't either and neither do most others.

jpeg vs tiff (http://www.kenrockwell.com/tech/raw.htm)

[shu246]

JPG is a compressed (shorthand) write of the data. Whenever you alter the picture in (e.g.) photoshop, some detail is lost in the translation.

Some pictures I expect to tweek in photoshop and then print on 11 x 17 paper. For these I use TIFF. Everything else is JPG.

[jaykinghorn]

Wingnut,

Shoot JPEGs in camera then save the best images out as TIFFs.

Jay Kinghorn
RGB Imaging

[JavaDan]

For anything up to and including 8.5 X 11 you will be just fine with jpg-fine. I've blown up those resolutions to 8.5X11 borderless with stunning detail. My photo software has no issues with them. Although I haven't done one yet, I suspect 16X20 would be acceptable. You're probably waisting your storage with tiff's if you don't need to go that large. If you plan on doing a lot of cropping and collages then perhaps tiff's would be better, but otherwise - stick with jpeg and save the space.

[Wingnut]

Thanks. I'll stick with jpg.
-Wingnut

[arghman]

btw, is the TIFF mode on FZ cameras 8-bit color or 12-bit? That would be the only reason I would use TIFF mode; in high-contrast situations the camera has to throw away some dynamic range because JPEG is an 8-bit format.

[jaykinghorn]

Jason,

Your comment about the 8-bit throwing away dynamic range is technically not correct. The difference between 8 and 12 or 16 bits is not that the higher bit images contain brighter whites or darker shadows. They share the same white point and black point. What differs is that 8-bit images slice that range into 256 steps. High bit images slice that tonal range into thousands of steps giving you more data for editing.

Another note about the JPEG vs. TIFF. I never advise my clients to shoot TIFF in camera, but I always advise them to save their files after making adjustments as either TIFFs or PSDs. Every time a JPEG is saved it loses detail. The first save from the CCD to the memory card is not something to worry about, but subsequent saves of the same image will degrade it. This is not the case with lossless formats such as TIFF or PSD.

Go ahead and shoot JPEGs in camera (there's really no reason to shoot TIFF), but save your important images as a TIFF or PSD.

Best regards,
Jay Kinghorn
RGB Imaging

[arghman]

Your comment about the 8-bit throwing away dynamic range is technically not correct. The difference between 8 and 12 or 16 bits is not that the higher bit images contain brighter whites or darker shadows.interesting... yeah, I did read somewhere after I posted that jpeg format allows for nonlinear quantization... how can we tell if the fz20 is taking advantage of this? do you know if there's a good technical summary of the jpeg format somewhere? is the nonlinearity (companding, is that the right word?) different for each 8x8 block or is it global for the whole jpeg file?

[arghman]

hmm, maybe "dynamic range" was the wrong phrase. Here's what I meant to say: suppose your subject has a very dark black area, with color variability in the black, and a very light white area, with color variability in the white. Then if you had an 8-bit digitizer and a 12-bit digitizer that were both linear (equal steps in each color's brightness from dark to light), then the 12-bit would pick up more detail. I'm under the impression that the fz20 (like most digital cameras) uses 12-bit digitizer but then has to post-process it to fit into jpeg which is 8-bit. So each original pixel has 4096 different values (each for R, G, B) and you have to squish that information into only 256 values or "bins" (also each for R, G, B). If you're not smart about it, all the fine shades of black get lumped together in the same bin, ditto for all the fine shades of white.

The kenrockwell.com link that genece posted, is the one that mentioned the nonlinearity (e.g. if you have a picture which has 12-bits = 4096 different values for each RGB pixel, then you have a chance to be smart and use different size "bins"... if there's lots of black maybe the lower-end "bins" of the JPG's 256 values discriminate among the darker shades, and the remaining bins take care of the rest of the colors.)

sorry if this is getting too technical for those of you who just like to shoot pictures...

[jaykinghorn]

suppose your subject has a very dark black area, with color variability in the black, and a very light white area, with color variability in the white. Then if you had an 8-bit digitizer and a 12-bit digitizer that were both linear (equal steps in each color's brightness from dark to light), then the 12-bit would pick up more detail. I'm under the impression that the fz20 (like most digital cameras) uses 12-bit digitizer but then has to post-process it to fit into jpeg which is 8-bit. So each original pixel has 4096 different values (each for R, G, B) and you have to squish that information into only 256 values or "bins" (also each for R, G, B). If you're not smart about it, all the fine shades of black get lumped together in the same bin, ditto for all the fine shades of white."

The problem is that the way that cameras record tonal values isn't in even brightness steps as you describe it. Cameras capture light and store it prior to processing with a linear (1.0) gamma. This devotes significantly fewer bins to the shadows than to the highlights. Half of the values are devoted to storing the brightest f-stop, while the rest of the information gets lumped into the remaining values. This is why accurate exposure is critical for maintaining shadow detail.

Best regards,
Jay Kinghorn
RGB Imaging

[hoffer]

I've been reading for several years that if you save changes to a JPG with the same "file name" it will degrade and that, after making changes, if you "save as" and use a different name there isn't any loss in quality.

Is this not correct?

Thanks, Rick

[pixelator]

Every time you save in JPEG, the image is degraded. Save in tiff until the image is definitive, and then save it as Jpeg

[pixelator]

The problem is that the way that cameras record tonal values isn't in even brightness steps as you describe it. Cameras capture light and store it prior to processing with a linear (1.0) gamma. This devotes significantly fewer bins to the shadows than to the highlights. Half of the values are devoted to storing the brightest f-stop, while the rest of the information gets lumped into the remaining values. This is why accurate exposure is critical for maintaining shadow detail.

Best regards,
Jay Kinghorn
RGB Imaging

Would be a RAW mode...

[jaykinghorn]

Currently the best solution would be to use raw mode. Most of the top DSLR cameras, along with an increasing number of point and shoot cameras are able to shoot in raw mode. This still doesn't solve our problem of not having enough values in the shadows as the analog light values enter the digital world. At this point though, there's not much we can do about the shadow values other than to be judicious about exposure and shoot RAW on cameras that support it.

Jay

[arghman]

the solution would be a RAW mode...Well, not really, I think jaykinghorn's point was that if your camera's ADC (where it converts analog information from the CCD to digital values to be processed) is linear, then half of its possible values are used to record the brightest portions of the image, and the darker portions get short shrift. If you had a nonlinear (e.g. logarithmic) A/D it might be better able to match the CCD's range of light/dark with people's nonlinear perception of light/dark. (I'm not aware of any nonlinear A/Ds sold commercially; you could combine a linear A/D with a logarithmic preamplifier but that would be more expensive and probably has other disadvantages.)

People (and animals) tend to perceive things on a logarithmic scale; e.g. if you had a part of an image with 5 shades of grey that looked equally spaced to the eye, but you took a look at the raw values (I'm pretty sure this is what "RAW" mode means, the raw unprocessed info from the A/D converter) measured by the camera, it might show them as 30, 60, 120, 240, and 480. The camera's A/D has much finer discrimination in the brighter areas (1 A/D count out of 480 is 0.2%) and much coarser discrimination in the darker areas (1 A/D count out of 30 is 3.3%).

So by the time you've got digital bits into the camera's processor, it's "too late", you've already thrown away some detail in the darker tones. My question was more about whether JPG's 8 bits do a good job in preserving the 12-bit data.

This is why accurate exposure is critical for maintaining shadow detail.So ideally, does that mean you'd want to make the exposure bright enough so it almost saturates some pixels (to make best use of the A/D), and then if it looks washed out, darken it in Photoshop?
In other words, overexposed images are easier to "fix" than underexposed as far as preserving detail, as long as they don't saturate the camera? (again, sorry about the length of this post)

[jaykinghorn]

Jason,

Very well put. That was a great explanation. Obviously you understand the nitty gritty of encoding better than I do.

You are correct that you want to overexpose to the point of almost oversaturating some pixels. The thought behind this is that you are trying to get your shadows into an area of the gamma that contains more bits. Bruce Fraser's excellent book Real World Camera Raw does a great job of detailing exactly why. Also, shooting RAW, you are often able to pull in overexposed areas of 1/2 to 2/3 of a stop. You have more headroom on the highlight side than the shadow. Plus, it's a lot easier to set a new shadow point in Levels, Curves or Camera Raw than it is to try and tease out underexposed shadow detail. The old saw about "expose for the shadows and develop for the highlights" has been flipped upside down for digital.

Cheers,

Jay Kinghorn
RGB Imaging