Image compression, we all seem familiar with the term but do not necessarily understand what it does to compromise an image. Basically what compression allows is to effectively reduce image data so that it has a smaller data occupancy. The same image can have many compression ratios applied, depending on the amount will increase the damage caused to the image, in effect you are reducing the data content so it can either be stored or transmitted more efficiently, such as over the internet.
Image compression is sometimes referred to as lossy or lossless. Lossless compression is sometimes preferred for artificial images such as technical drawings, icons or comics. This is because lossy compression methods, especially when used at low bit rates, introduce compression artefacts which can be seen. Lossless compression methods may also be preferred for high value content, such as medical imagery or image scans made for archival purposes. Lossy methods are especially suitable for natural images such as photos in applications where minor (sometimes imperceptible) loss of image data is acceptable to achieve a substantial reduction in the size of the image. It should be noted when you compress an image, whatever data is removed then it can not be retrieved at a later date. There are many ways of compressing images and the user can dial in the amount of compression, so it can be a particular size that is needed for a particular application, such as print or web based materials.
However for film based applications it is recognised that any banding or artifacts (jaggies/aliasing problems) will be seen when blown up on a larger scale when projected. Therefore the DI has always been carried out at the highest possible image quality available, which has been standardised at 10 bit log, although 16 bit uncompressed tiffs are beginning to gain popularity.
Re sampling and adjustment in any form uses maths on an image and degrades or damages it - plus saving it within certain formats will compress the image, good enough for the web but may not be good enough for other applications. Especially for colour correction of FX work.
Adding a very high compression algorithm is plainly obvious in the image below on the left, you will see extreme artefacts and blocking gradation within the image also the detailing disappears within the image, this is quite an extreme example. The image on the right is demonstrating the anomalies between an uncompressed and compressed image, which may not be obvious purely looking at the separate images but may be seen when performing further processes on it.
Obvious points of compression are in the tree itself but if you look closely there has been a banding applied to the sky which is not that noticeable when you look at the compressed image above (it is there if you look closely).
Ultimately an image can look pretty good and these artifacts may not be noticeable until you perform an intense correction or a high level composite, only then the image may break down and show these anomalies, especially if you are using a combination of compressed and uncompressed images. It should be noted even using 8 bit with 10 bit images together can be enough to see problems.
Below the image on the left is uncompressed with an extremely harsh grade applied and the image on the right has the same grade applied to a compressed image. Hopefully this demonstrates how compression can cause the image irreparable damage.
During the Digital Intermediate process, it should be common practice to work with the highest quality images available, which will ultimately give the highest quality look and output for the intended media. Obviously there are exceptions because sometimes material is only available at a smaller resolution or with compression - knowing how to deal with it to get the best from the material takes some considerable testing, there is no single solution, if only it were that easy!