(note: I am currently moving and half on vacation. I am separated from my normal environment until August 15th. I can’t post as many examples as I like, and in this article, I am using a couple recycled images that are great examples of dramatic Sagelight edits on 8-bit JPEGs — highly compressed ones, at that!)
The other day on the discussion board (www.sagelightforum.com), the issue of working with images that start as JPEG vs RAW format came up, as it does quite often.
There’s definitely one or more schools of thought about that — no one of them right or wrong, as this – as well as most issues with image editing – should be and are really up to the person behind the image and his or her own preferences.
In my case, for example, I set my camera to shoot in RAW and JPEG both. I edit the JPEG nearly exclusively (see the next section about that), and turn to the RAW image when there is something about the image the JPEG format (or camera in its JPEG conversion) couldn’t handle. One of the biggest issues, for example is highlights. Highlights get blown out in quite a few pictures (as well as shadows), and it really is amazing how well the highlights can be recovered from a RAW image (shadows, too). In a JPEG image, if the highlights are gone, they’re pretty-much gone, although some highlight recovery is possible. One way to tell is that if the right edge of the histogram is peaked and filled up from the left, then the highlights are probably unrecoverable in a JPEG. This is not so with a RAW image because of the bitwise resolution, which allows it to store as much as an entire JPEG spectrum in the equivalent of 1 8-bit value of a JPEG.
In this post, I don’t talk about the advantages of using RAW. These advantages are clear for those that prefer those over the advantages working with a JPEG presents. It’s an individual choice, and I talk about the advantages of RAW in other posts. In this post, the discussion is purely about why working with the JPEG not only has its own advantages, but — in most cases, with just a couple exceptions — makes no difference (or little enough to be an issue) in the final image, except for those looking to keep the maximum quality to their image, such as professionals and serious hobbyists.
note: I will balance this post soon by posting the Advantages of using RAW.
The introduction picture is a great example of an image that started in 8-bits per-channel. This image (directly above) is the original image. This started as a fairly lossy (i.e. low-quality) JPEG in 8-bits per-channel. You can see how dramatic the difference is with the image, and how it suffered very little for being in 8-bits per-channel. Would the RAW, uncompressed version come out better? Maybe and probably, but sometimes this is the difference between spending 3 minutes on an image vs. 10-20, and also the difference between working on an image for fun and not working on it at all, at least for those that don’t like the extra steps one usually needs to take to start with a RAW image. Again, it’s a personal choice, and this post is oriented towards those who prefer to work in JPEG and have been told it’s a bad idea – it isn’t and, in fact, working with JPEGs can sometimes be a great idea!
It’s the Quality of the JPEG and not the 8-bits Per-Channel That is Usually the Culprit With Image Degradation
When I do work with the JPEG, it is important to note that the settings on the camera are at the largest size/lowest compression ratio. With Sagelight, for example, if you choose the ‘best’ setting when saving out a JPEG, it is nearly lossless. When you start moving to lower numbers, then you can start to see the difference. The trick here is that the human eye really can’t perceive the things that degrade the image for editing, which is why it is always best to save your image at the ‘best’ setting as a JPEG, except for that one final version you want to put out to the web or your personal collection. For working with images, I advise to save out to .TIFF format, as it is lossless as well as 16-bits per-channel.
Most of the time, the degradation problems with working with images that start as JPEGs are because of the quality-level issues and not the 8-bits per-channel. 8-bits per-channel does have its own problems, as does the maximum quality settings with JPEG, but, besides the exceptions mentioned above, this usually isn’t a problem.
JPEG is an already-enhanced edit of your image
I often talk about how there should be a minimum of two stages when editing RAW images. The first stage it to get it into minimal shape to do the more aesthetic, general editing. RAW’s start off very different than the image you want to edit, and there are many ‘preparatory’ steps taken with RAW, such as the initial brightening, sharpening, NR, etc. After this is completed, then you can edit the image in more detail. It’s not good for a RAW image to perform everything you want to do it in one session, which is why most programs that handle RAW images either do these things automatically for you, or give you some control as a pre-step to general editing.
When the image is stored as a JPEG on the camera, it is first lens corrected, brightened, noise-reduced, possibly saturated, and whatever other things the camera manufacturer has decided works well for most images. All of this is done in the native bitwise resolution (at least for cameras that support RAW, but probably ALL cameras anyway), then saved for you to load later. In reality, you’re never starting with an image that has solely been edited as an 8-bit image.
Extending that, even when you load a JPEG that is 8-bits per-channel (as are all standard JPEG images), the 8-bit factor does not stay as much of an issue for more than one step in the editing process.
Working in 8-bits per-channel vs. working with an image original stored/retrieved as 8-bits per-channel
Since I started Sagelight, and more particularly in the last year, I have run into a number of things I consider to be old-school thinking and yet are still common issues. For example, the whole destructive vs. non-destructive issue I’ve been writing about lately, in part, comes from the notion that all ‘destructive’ (i.e. changes to your image) is bad, when it can be healthy for your image and is necessary if you want to do other and more interesting things to your image than adjust it with a small and finite set of controls. A few years ago, when many editors were destructive to your image in the wrong way, instead of a way that preserves quality; and there were so many less things you’d want to do that require the changes to your image be applied, the idea of a 100% destructive editor made much more sense, when now it is actually a limiting factor.
Another one of those notions that comes from just a few years ago — that made sense at the time, and has been mischaracterized lately — is the idea of working with an ‘8-bit’ image or ‘in 8-bits’. The difference between working with an ‘8-bit image’ (which is fine) and working ‘in 8-bits’ is a very large and important difference.
Working ‘in 8-bits’, which means working in an editor that keeps your image stored as 8-bit per-channel and/or performs operations at 8-bit per-channel, is not good for your image and will eventually cause visible problems.
However, working with an ‘8-bit image’, one that originates in 8-bits, such as the JPEG retrieved from your camera is just fine, as long as it is an editor that isn’t 8-bits itself.
Once your image is put into Sagelight’s memory, it is stored in at least 16 bits-per-channel, and most operations are in 32- or 64- bits per-channel. This means that, while there is an initial resolution of only 8-bits, subsequent operations are in a much higher quality, which preserves a lot more of the original information and details as you continue to edit the image. Some tend to confuse the idea of an 8-bit image with working in 8-bits, which is not the case, and is an entirely different issue.
In addition to the higher working resolution, many functions use blending, which is more of a modern-day commonality than just a few years ago. When you use the Fill Light, Light Blender, Definition, Focus, Noise Reduction, Smart Light, Bokeh, and most other functions, blending is performed on your image at some level. This means that Sagelight – to some degree or another – blends the pixels in your image, either directly, or through the colors/light only, or feathered percentage (i.e. it scales down on the borders of light/color areas very smoothly), adding and creating higher resolution in your image.
A number of functions also help you correct issues in your image that aid in creating a higher resolution per-channel in your image. Smoothing, Sagelight Noise Reduction, Soft Light, Skin Smoother, Bokeh/Lens Blur, and quite a few other functions will add more resolution to your image as you apply them.
For example, if you take an 8-bit per-channel image and sharpen it with the Unsharp Mask, in most images you will see the gradation lines created by the sharpening. However, if you apply a round of noise reduction to the image and then sharpen it, you will typically see no gradation lines at all and it will be clearer (and higher-quality in the smooth areas) than if you sharpened a raw 16-bit image without the NR pass.
Cloning and the Wire Worm
Another form of correction is cloning and the wire worm. This will help correct little ‘hot spots’ of trouble in your image manually. In some images that start as 8-bits per-channel, some highlights and shadows can be completely blown out or flat. Other areas can be splotchy. The clone brush or wire worm can be very useful to address these areas by borrowing more intact areas of the image.
Resizing your image is also a form of effectively adding resolution to your image. Upsizing or downsizing can help here. For the most part, I am speaking of downsizing your image, and will talk about the upsizing concept in some other post.
Most cameras these days take rather large images. When you publish them to their final result, they are often sized down, even if just for viewing. When you resize your image to a smaller size, many of the problems that are in the image – starting at any bits per-channel – merge and suddenly disappear. In many cases, you can take a very noisy image, resize it to a smaller size, and the noise is suddenly gone.
two notes here:
- Sometimes sizing down creates its own little noise — use the Sagelight NR to get rid of this noise. In many cases, a small NR pass on the final image can help it out.
- You can control this a little, as well. I recommend Lanzcos 3 for downsizing; however, it is so effective, it shows more of the original noise than any other methods. Try blinear or bicubic for resizing — bicubic resizing is very sharp in Sagelight, but also helps blend image problems. Blinear resizing’s only drawback is that it can create moire patterns. Bicubic is very nice also, being much softer than Lanzcos or Bilinear, and avoiding moire issues.
One More Example
Here is another before and after of an image that was loaded in Sagelight as an 8-bit image, but otherwise editing as a 16-,32-or 64-bit image internally:
Original Image. Not bad already. Let’s see what we can do with it:
As you can see, I was able to bring out a lot of light and definition without the image suffering at all. If I were to edit it a little more, I might smooth the highlights in the clouds a little, but that’s a personal choice (that is, I might have overdone the highlights a little here, but that was my editing, not the limitations set by the 8-bit per-channel source, since the highlights weren’t there in the first place), not to mention a reflection after having edited the picture well over a year ago.
There is a very large difference between working with an image ‘in 8-bits per-channel’ and working with an image loaded from storage that is 8-bits per-channel itself. Working with these types of images in Sagelight is fine, and can have many advantages over the higher quality when working with RAW images. Those advantages include the ease of editing by avoiding the typical RAW preparatory steps, and the more casual approach it allows us to take.
My personal philosophy is to work with the JPEG unless there is a compelling reason to work with the RAW (such as blown out highlights). As long as the image is stored in the highest quality and lowest compression, otherwise working with an 8-bit JPEG often comes out every bit as nice, or so close as to not matter, as the RAW result.
Without taking sides on the RAW vs. JPEG debate, and also keeping in mind that using RAW is always going to give you the best quality; using JPEG most of the time can be fun and direct, avoiding a lot of time spent with the RAW original.
With the advanced tools and strategies for modern-day images, the idea of working with an 8-bit image is much more compelling. Tools such as the Sagelight Noise Reduction, Smoothing, Soft Glow, Blending Functions (such as Light Blender, Smart Light, etc.) can literally add bitwise quality to your image. You can also fix up isolated problem areas with many tools, such as the Clone Brush or provided Wire Worm plugin.
Ultimatelt, it is – and should be – a personal choice whether to use RAW or JPEG. When you see posts on discussion boards, or your peers tell you that editing with the original JPEG is unworthy, this, in my mind, can be either just differences of opinion or old-school thinking that does not take into account a lot of the advanced toolsets available out there, such as many of the tools inside of Sagelight.
The two examples presented here were created in just 2-3 minutes each, and the fact they started in 8-bits per-channel did not cause any problems. Most before-and-after images you see on the blog posts and in Sagelight have started as 8-bits per-channel, as these tend to be the sources I get. I do have to throw away some here and there for lack of quality where the RAW Image might have worked, but, overall, it’s been worth it to primarily edit the 8-bit per-channel source.
The main reason for this is two-fold:
- The image is immediately converted to 16-bits per-channel, and all operations are usually in at least 32-bits per-channel
- The tools in Sagelight keep the quality high and often actually add bitwise resolution to the image, to the point where it’s only the odd/exceptional case where I need to refer to the RAW.
It’s always good to have the RAW source around for those tough images, and, from a technical point of view, editing a RAW image is higher quality than a JPEG image. But often, if doesn’t really matter. A while back, I likened this issue to those who prefer vinyl records over digital. I absolutely respect that and support a lot of RAW functionality in Sagelight. Vinyl records, indeed, have higher quality than digital music, as long as they are kept scratch-free and in pure condition. Much like digital music, as long as it is in CD quality (i.e. 8-bits, but at the highest-quality, lowest-compression JPEG), then that is just fine for most of us, too.