Byte sized: JPEGmini claims no loss of perceptual quality, up to 80% smaller files
Most of us think of image compression as a necessary evil. It makes our files more manageable in terms of size, but reduces the quality of our images and can undo the incremental benefits of buying more pixels and better lenses. If offered the choice between more or less image compression, I suspect that most photographers would always go for less. Hence, the idea of buying a piece of software that aims to reduce the size of JPEG files by up to 80% might seem a little crazy. But that is exactly what Beamr, the company behind the JPEGmini application, is offering.
Introduced in 2011, JPEGmini acts as a standalone product or as a plug-in for Lightroom and is a compression optimizer that takes in existing JPEG files and makes them smaller – without reducing the quality of the image, the company claims. The idea is to save space on hard drives, external storage devices, make websites run more quickly, deliver more manageable file sizes to clients and help reduce spending on cloud storage. We've read up on it and written about its desktop and mobile applications briefly, but Senior DPReview Contributor Damien Demolder recently had the chance to sit down with the company's Chief Technology Officer to find out more about how it works.
How it works
|Dror Gill, CTO and VP of Beamr, the company behind JPEGmini|
In an interview, Beamr's CTO Dror Gill explains how JPEGmini works and how the company measures its claimed 'no change' in image quality.
'JPEGmini works with standard JPEGs. The input is a standard JPEG and the output is a standard JPEG. We recompress that standard JPEG photo by up to 80%, and the resolution remains the same and the perceptual quality of the image remains the same. When we talk about 'perceptual image quality' we mean that if you took this photo and viewed it on your screen at Actual Pixels, or 100% magnification, and compared it to the original you wouldn't be able to determine which was the original and which one was the optimized. That's what we call 'perceptually identical' to the original.'
I wanted to know who the 'you' was in that qualification – as the opinions of a general consumer, a photographer and a scientist will all be significantly different. Gill said that '99% of the population' wouldn't be able to tell the difference, including most photographers.
'Any JPEG compression introduces artefacts, but the question is,
are these artefacts visible by humans or not?'
'Most of our customers are professional photographers, and they have realised that the photos that they get out of JPEGmini are as good as the originals and that they can use them in the same situations and for the same uses. Of course, the JPEG process introduces artefacts that you don't find in the RAW file, so any JPEG produced by Photoshop or Lightroom will have artefacts, but our claim is that our processed image will look the same as the original JPEG and the compression will not introduce further artefacts. Any JPEG compression introduces artefacts, but the question is, are these artefacts visible by humans or not? We have developed a quality measure that gives us that answer with very high accuracy. This quality measure has much better correlation with human results than other scientific quality measures.'
The software works by analyzing the content of each image, and determining how much compression can be applied to each individual area. Images are broken down into tiles of a set number of pixels, and the degree of compression acceptable is assessed according to the level of information recorded in the tile. Gill wouldn't say how the tiles interact with each other, but we worked on the presumption that the tiles were about 150 pixels square.
If there isn't much data recorded the content can be compressed more than if a tile contains a lot of data, so the savings are made via a more flexible process than the usual global compression ratios that most software applications and cameras work with. The software produces compression level 'candidates' for each tile, which basically means it tries different levels and determines the maximum that can be achieved without loss of the information in the tile – and then that amount is applied.
Gill says camera manufacturers don't like to use a lot of compression because too many reviewers and customers think that image quality and the amount of detail in an image can be determined by the size of the file created, and that people associate smaller file sizes with lower levels of picture information. Camera brands, he says, don't want to produce files that are smaller than their competitors as some reviewers will immediately mark them down for it without studying the comparison images.
'what we do is take that image and determine
what is the exact optimal level of compression for that particular picture'
Cameras don't have any mechanism for evaluating the content of the image either, he says, so the compression has to be global and to err on the safe side. 'This results in a relatively large JPEG,' says Gill, 'but what we do is take that image and determine what is the exact optimal level of compression for that particular picture. Some images are more easily compressed than others – some have very delicate textures and smoothly varying color gradients, and for those you need to use high quality settings. If the content is mainly smooth surfaces and busy backgrounds, that you can't tell if they are degraded or not, you can use a higher compression ratio.'
Gill says that out-of-focus backgrounds can be compressed more than focused areas, as the software analysis works by detecting the amount of detail and information present. This brings up the question of whether a poor lens will be made to look worse by the compression compared to the same area captured by a sharp lens, but Gill maintains that the difference wouldn't show. Tests, I suppose, will give us the sure answer to that.
If you view the optimized images at 800% Gill admits that you would see the differences, but at normal viewing and for normal use you won't. 'These optimised files are designed to be viewed at 100% and to be printed. In print it is even harder to see the differences than on screen.'
'the inefficiency of normal JPEG compression pollutes the environment'
The whole idea of JPEGmini, Gill explains, is to save space on laptops, hard disks, online and in external storage. 'There are a lot of terabytes wasted by files that are larger than they need to be. There is no point using bytes and bits that are not visible to humans. The industry is doing it all the time. Maybe we should calculate how many exabytes are being wasted every day – the inefficiency of normal JPEG compression pollutes the environment' he only half-jokes.
Gill's father is Aaron Gill, who was one of the chief scientists who worked on the original JPEG standard in the 1980s. I ask how he feels about his son tampering with the way JPEGs are created. 'At first he was sceptical and asked me what I was doing getting mixed up with this company that wants to reduce file sizes, but after he tried it I think he was proud of me.'
Trying it out
JPEGmini supports JPEG files up to 28MP, while its JPEGmini pro and JPEGmini Server siblings support up to 60MP images. To give an idea of what JPEGmini does, I ran a 25.45MB Raw file through Lightroom and exported a 'best quality' JPEG of 10.12MB. After being exported again via the JPEGmini plug-in the file was compressed to 2.66MB, and still measured the same 4608x3456 (16MP) pixels it did originally - so the JPEGmini file is a quarter of the size of the normal JPEG.
The software still makes considerable savings even if you don't usually convert your images using the best quality settings. For comparison, that Raw file exported as a JPEG at 80% quality in Lightroom (not using JPEGmini) resulted in a 4.8MB file. The 2.6MB JPEGmini file is just over half the size.
Although photographers might like the idea of saving space most are not interested in doing so at the cost of quality, and frankly I think most of us struggle to believe that such a dramatic file size reduction can be achieved without any detrimental effect on the content of the picture.
|Normal JPEG exported from Lightroom at best quality||JPEG exported from Lightroom via the JPEGmini plug-in|
In my very brief tests I have been able to see slight differences in levels of micro contrast and the amount of very fine texture that is resolved when the images I used were viewed at 100% on screen. More tests will be required to see exactly what is lost and what is at stake, and I'm compelled to make those tests by the carrot of saving a massive amount of space in storage and by the prospect of having a website with large images that runs quickly. At this stage I can say that in the image I tested the plug-in with tiny differences could be seen when the images were compared at 100%, but at print size (33%) the differences were certainly not apparent.
|Normal JPEG converted from Raw at quality 11 - 2.2MB||JPEG further compressed via JPEGmini app - 980KB|
If you can't wait for the results of my testing you can download the $19.99 standard standalone version of JPEGmini for a free trial. JPEGmini Pro costs $149 but can work with images of up to 60MP, is up to 8x quicker and comes with the Lightroom plug-in option as well as the standalone application. At the moment however, JPEGmini only accepts JPEG files. That means even using the Lightroom plug-in, a Raw file must first be converted to JPEG to then be re-saved as a smaller JPEG by the application.
For more information visit the JPEGmini website.
Aug 24, 2016
Jul 6, 2016
Jun 28, 2016
May 27, 2016
- Canon EOS M58.8%
- Panasonic G85/G803.3%
- Panasonic FZ2500/FZ20001.9%
- Panasonic LX10/LX151.2%
- Panasonic GH5 development3.6%
- Sony a99 II15.9%
- Nikon KeyMission 170 and 801.0%
- Fujifilm GFX 50S development28.3%
- Olympus E-M1 II development18.7%
- Olympus E-PL80.1%
- Olympus 25mm F1.2 Pro1.5%
- Olympus 12-100mm F4 IS Pro1.9%
- Olympus 30mm F3.5 Macro0.1%
- Sigma 85mm F1.4 Art3.6%
- Sigma 12-24mm F4 Art2.6%
- Sigma 500mm F4 DG OS HSM Sport2.4%
- YI M12.2%
- GoPro Hero50.8%
- GoPro Karma drone2.2%