I'm not sure what type of edits you performed in the first place that you're undoing and how you're undoing them. If it's anything other than using the Ctrl-Z or going back using the history in photoshop, you're not really undoing your edits. You're adding more edits to give the effect of undoing your edits.
Also, directly going from RAW to jpeg can potentially give you better quality than going from RAW to Tiff to jpeg depending on the compression setting that's used in the raw decoder and the setting used in the tif to jpeg encoder.
Herb also made a good point that certain Tiff encoders will apply compression depending on the settings that you use. Only uncompressed tiffs and (maybe) losslessly compressed will maintain its original information that it received from the raw converter.
beewee wrote:
I'm not sure what type of edits you performed in the first place that you're undoing and how you're undoing them. If it's anything other than using the Ctrl-Z or going back using the history in photoshop, you're not really undoing your edits. You're adding more edits to give the effect of undoing your edits.
Yea, I just used the undo button, Ctrl-Z. Thus everything that was changed was undone.
beewee wrote:
Also, directly going from RAW to jpeg can potentially give you better quality than going from RAW to Tiff to jpeg depending on the compression setting that's used in the raw decoder and the setting used in the tif to jpeg encoder.
This may be the case, but again, both images should be the same. My problem is the difference between the images. Even if the images were degraded for the reason you suggest above then they should both have degraded equally.
beewee wrote:
Herb also made a good point that certain Tiff encoders will apply compression depending on the settings that you use. Only uncompressed tiffs and (maybe) losslessly compressed will maintain its original information that it received from the raw converter.
True, but again this would not account for the difference in the images.
I can think of no good reason why the images should differ.
beewee wrote: Also, directly going from RAW to jpeg can potentially give you better quality than going from RAW to Tiff to jpeg depending on the compression setting that's used in the raw decoder and the setting used in the tif to jpeg encoder.
michael49 wrote:
This may be the case, but again, both images should be the same. My problem is the difference between the images. Even if the images were degraded for the reason you suggest above then they should both have degraded equally.
beewee wrote: Herb also made a good point that certain Tiff encoders will apply compression depending on the settings that you use. Only uncompressed tiffs and (maybe) losslessly compressed will maintain its original information that it received from the raw converter.
michael49 wrote:
True, but again this would not account for the difference in the images.
I can think of no good reason why the images should differ.
I feel like I'm repeating myself but not all jpeg encoders are made equal and even with the same jpeg encoder, if you use different encoding parameters, you can expect to have very different results. Google has a very good lecture on lossy compression. Perhaps this might help give you a better understanding of jpeg compression.
I just skimmed the thread as this really all boils down to one thing: ALL edits throw away some data, regardless of what format you use. Every time you perform an edit in a pixel editor (PS, PSP, etc), you are throwing away some data. When you save to JPEG, the saving process throws away yet more data while saving to TIFF does not.
This is why such a big deal is made about the non-destructive editing available to you in RAW converters. Some image-wide changes can be made at the time of converting to an RGB format which does not/minimizes the kind of degradation you are seeing.
michael49 wrote:
The only "permanent" change I made to each image was straightening, which I did to each image. I thus saved the image after doing this.
Straightening a file moves the pixels around (very untechnical explanation, but short ). Imaging programs have to have some anti-aliasing built in to the straightening or else you'd get jaggies on straight lines. What you're seeing in your straightened files is a little softening. This is perfectly normal and is to be expected with a rotated file. It has nothing to do with TIFF format; IOW, the loss of detail is due to the editing operation, not the save into TIFF. Use some sharpening after straightening.
michael49: If by "straightening" you mean rotating -- there's your answer. Otherwise you or your software is NOT undoing the edits. BTW -- "TIFF" includes a LOT of formats. Typically around here when people say TIFF they mean plain ol' RGB TIF with a 16 bit depth per color. Really no point in using TIFF in photo processing unless that is what you are using.
Chuck: it's not Lab, it's YCrCb aka YCC for short. Same principle, the fundamental compression behind JPEG is separating luminance from chrominance as the human eye is roughly twice as sensitive to luminance variations than chrominance variations so an increased level of compression can be had by throwing away twice as much chroma as luma info for the same apparent quality, plus of course the 8x8 coefficient transform. What is not generally discussed, but of direct impact is the quality of the encoding in the intermediate parts of the algorithm and the rounding errors introduced. The "quality" setting is randomly implemented. There is a big difference in what software is used to encode JPEGs. Anyone doubting this: simply take a large JPEG direct from the camera, load it into PS, save it as a copy without any other processing in PS with Q level 10, then pixel peep. After finding no differences, go look at the file size difference .
Folks, there is such a thing as lossless JPEG, we discuss and use it a lot in medical imaging. You can render, and resave a JPEG without any loss of quality if the transform operations and saving are done with such a bit depth that any errors introduced are sufficiently beyond the bit depth of the display hardware that it is mathematically proveable that there cannot be any noticeable difference.
It just kills me how backward things are in this world. If the the freakin' cameras would just encode the images in a wide color space at 16 bit depth, far in excess of the sensor technology, and the world would deal with JPEG 2000, half of the discussions in this forum would simply go away.
Edited by christo™ on Feb 12, 2008 at 03:55 AM GMT
michael49 wrote:
The only "permanent" change I made to each image was straightening, which I did to each image. I thus saved the image after doing this.
I must have missed this important bit. When you "straighten" an image, you're essentially doing a 2D transformation using a 2x2 rotation matrix. The only time that you will not see any change in sharpness is if you do a full 90 or 180 degree rotation. Otherwise pixels must be interpolated to fit into their new location. There are a bunch of ways to do this but none of them are perfect.
You can either use the "nearest neighbour" algorithm where if you imagine 2 grids of pixels, one slightly rotated, for each new location of the pixel grid, you pick the colour of the pixel from the nearest old pixel grid.
Other algorithms use some forms of interpolation, averaging, distance weighted averaging, etc.. You can read those up if you're interested.
he did it to both images. i made the assumption that he did transformation using a dialog box and entered the degrees of rotation. if he did it freehand, then all bets are off.
Herb...
beewee wrote:
I must have missed this important bit. When you "straighten" an image, you're essentially doing a 2D transformation using a 2x2 rotation matrix. The only time that you will not see any change in sharpness is if you do a full 90 or 180 degree rotation. Otherwise pixels must be interpolated to fit into their new location. There are a bunch of ways to do this but none of them are perfect.
There's tiff and then there's tiff. If you convert from 12- or 14-bit raw file to 8-bit tiff then you've lost almost as much data converting from raw to jpg. If you want to preserve the data as long as possible in order to minimise any damage from significant editing of colours and tones then you need 16-bit tiff. After the editing is all done then dumb it down to 8-bit. At that point you might as well go straight to jpeg because there won't be any more editing.
tiff is only lossless in that if you open it and resave it and open that one then you will have an exacly identical image whereas if you do that with a jpg then it will be re-compressed and hence altered. Neither format is lossless when you start editing the image data and doing damage to it.
Both formats use integers to store the image data and so there can be no in-between values. The more integers you've got to work with (and actually use) the more you can do without losing too much significant information simply because each increment in the integer value has less meaning. Throwing away much of the data to convert it to 8-bit before editing can result is major quality loss. That's why it pays to use 16-bit tiff instead of 8-bit tiff at least until the major editing is out of the way.
Apart from 8-bit vs 16-bit, jpg files lose out even more again because they keep compressing the data by discarding fine details every time you re-save the file. A high quality low compression high resolution jpg can withstand several saves without appreciable deterioration but significant 8-bit editing can ruin it.
1. 16 bpcpp (bits per color per pixel) rendered from RAW is the way to best avoid digital artifacts in edited images such as "banding." For sunsets and "green grasshopper on new grass", it's the only way to go.
2. The quality loss due to JPEG within a few generations of edit / save is more due to 8 bpcpp storage rather than the JPEG compression if you leave the Q factor in the JPEG compression at 10 or higher (in PS anyway.)
3. There isn't really any noticeable image deterioration in loading a large/fine JPEG and just saving it with Q 10 with PS. If the image has deteriorated, it's because the edits caused issued due to the low bits per pixel count.
). Imaging programs have to have some anti-aliasing built in to the straightening or else you'd get jaggies on straight lines. What you're seeing in your straightened files is a little softening. This is perfectly normal and is to be expected with a rotated file. It has nothing to do with TIFF format; IOW, the loss of detail is due to the editing operation, not the save into TIFF. Use some sharpening after straightening. 
.