LZW compression is quite common and indeed lossless. Common enough that I wouldn't say it's not standard (but who is setting the standard, here?) and JPEG compression in TIFF images happens enough that I would never say that no TIFF is ever lossless.Chris Peterson wrote:Standard TIFF files don't use JPEG compression. If they use compression at all, its lossless. The image I posted was made by taking the original TIFF image and recoding it using lossless JPEG, which is a valid form of JPEG compression. As a result, the original TIFF and my recoded JPEG are identical pixel-for-pixel (even thought the TIFF is around 25 times larger in terms of file size).geckzilla wrote:BTW, are you sure that a TIFF file which uses JPEG compression is lossless? I don't think this is possible. Even a TIFF which uses the least amount of JPEG compression is still slightly lossy.
Because TIFF images are lossless, they are commonly used by imagers to move data between different programs- especially between astronomical image processing apps that typically use FITS format, and Photoshop, which doesn't natively support FITS. Even the popular program FITS Liberator, used by many HST processors, converts FITS to TIFF for use in other programs.
TIFF supports two different internal compression schemes, both RLE variations and both lossless. While the standard is extendable, and some implementations might include JPEG or LZW compression (such as Photoshop), those would be non-standard file formats and there's no requirement that any given app would be able to read the file. When compatibility is required, files should be uncompressed (which is what most TIFF files are) or compressed only with one of the RLE algorithms required by the standard. The safest way to compress TIFF files is to ZIP (LZW) them after they are written.geckzilla wrote:LZW compression is quite common and indeed lossless. Common enough that I wouldn't say it's not standard (but who is setting the standard, here?) and JPEG compression in TIFF images happens enough that I would never say that no TIFF is ever lossless.
The JPEG specification allows for lossless compression. Some apps implement it, some not. However, even those that do not produce images visually indistinguishable from the original, and as you note, it would take a number of decode and re-encode cycles before any artifacts became visible.You may want to reconsider calling your JPEG lossless. Even at maximum quality and minimal compression, JPEG is still lossy. It may not be apparent but there are minute artifacts introduced. You could save and re-save it 10 times and begin to notice. A non-lossy format would not ever introduce artifacts.
PNG is good for basic graphics, but not generally a good choice for photographic images. Besides producing much larger files for typically insignificant (or even undetectable) improvements in quality, it's lousy dealing with color spaces in managed color environments (which is most these days). JPEG is really a near perfect format for most photographic images on the web, as long as the author takes a little care to select an appropriate degree of compression.geckzilla wrote:I think PNG would be a great choice. Of course, rural internet connections still aren't able to handle that kind of file size depending on how compressible the data is (quite compressible for this image, in fact! only 45k), but man, I can dream. 24 bits of lossless compression and an alpha channel to boot.
Chris, speaking of that, perhaps you could answer this question. When I covert an image from TIFF format to JPEG(using photoshop), the color space is set toChris Peterson wrote:PNG is good for basic graphics, but not generally a good choice for photographic images. Besides producing much larger files for typically insignificant (or even undetectable) improvements in quality, it's lousy dealing with color spaces in managed color environments (which is most these days). JPEG is really a near perfect format for most photographic images on the web, as long as the author takes a little care to select an appropriate degree of compression.
Chris Peterson wrote:Yes. These are bright objects (sunlit) and consequently the exposure times are short- less than a second. That's just not long enough for many stars to show up. It is possible to see the occasional bright star in the raw data, but only just above the noise floor. Take your camera out at night and shoot a properly exposed image of the Moon. You won't see any stars in the result.Swashbuckler wrote:
Why do we not see any background stars or other objects in photos like these? They're photos, and not filtered images for hydrogen or x-rays. Is it just the amount of reflective light pollution from the visible objects?
I always convert my web images to sRGB (if the sources aren't already in that space). Although the sRGB gamut is somewhat narrower than others, that's not usually a factor with images viewed in browsers, and sRGB is the assumed color space in most apps that aren't color managed. I figure using sRGB offers the best chance of the most people seeing something at least close to what I intend. Of course, it's also important to save the image with the color profile included, or results will be variable. This is a problem with many APOD images, which are frequently posted without color profiles. To be more precise, most of the submitters do provide images with color profiles, but the process used by the APOD editors to create the version that shows in the main page removes the color profile. When viewing APODs, it's important to click on the image to view the original if you want to see the correct color. (Would somebody chip in and buy those guys a copy of Photoshop? Last I heard they were using some 20 year old graphics package to produce the main page reductions/crops.)stephen63 wrote:Chris, speaking of that, perhaps you could answer this question. When I covert an image from TIFF format to JPEG(using photoshop), the color space is set to ICC Profile: sRGB IEC61966-2.1. Is this the working space most suitable for posting JPEGs on the web?
This varies a bit between color and grayscale images, but as a rule, JPEG can compress about 10 times without producing visible artifacts with most photographic images (a setting of about 70 in Photoshop). If the intent is an image for display only (no later editing), saving at this level will almost always produce very good results- effectively indistinguishable from an uncompressed image. BTW, this applies to typical terrestrial images, which are full of high frequency information. Astronomical images often have little high frequency content. These can compress by a factor of 20 or more even when the highest JPEG quality is selected, with even highly compressed images being indistinguishable from uncompressed ones.As far as compression goes, I don't compress at all unless there are file size restrictions, like here. Is there a better way than using trial and error on the amount of compression until file size limitation is met?
They don't want it or it would already be done.Chris Peterson wrote:(Would somebody chip in and buy those guys a copy of Photoshop? Last I heard they were using some 20 year old graphics package to produce the main page reductions/crops.)
There must be some newer simple graphics package that has no learning curve to speak of, and can be used for simple cropping and resizing while preserving color information. It's really unfortunate that the main page APOD images lack color profiles, and are therefore not being seen as intended by most viewers. That's pretty unscientific for a science-oriented site!geckzilla wrote:They don't want it or it would already be done.Chris Peterson wrote:(Would somebody chip in and buy those guys a copy of Photoshop? Last I heard they were using some 20 year old graphics package to produce the main page reductions/crops.)
