What Do You Lose When Compressing Images With JPEG?

We encounter and use JPEG files daily, but rarely do we give any thought to what the process entails. More importantly to us as photographers, what data do we lose when using this lossy image compression technique?

JPEG files are ubiquitous, and if you use any digital medium at all, you're bound to encounter them multiple times a day. They're responsible for being able to squeeze image data into files that are often a tenth of their uncompressed size but that seemingly have little to no loss of image quality. But JPEG is a lossy file format, so what is lost when you save a JPEG file?

Branch Education simplifies this complex topic that I never thought I'd understand in this short but informative video. They make it immensely accessible with visuals and animations that will allow you to understand topics such as "Discrete Cosine Transform" and "Quantization." On the way to understanding all this, they also explain some color theory and human anatomy.

The rods and cones of a human eye. Why JPEG works.

The process of JPEG image compression takes into account how our eyes work as well as how most photographs work to create files that are imperceptibly different from the original but that have significantly smaller file sizes.

Understanding the process and the five steps involved allows us as photographers to understand how our images are changed when saving them and also to understand just what we are trading in for smaller file sizes.

Chrominance versus luminance

You will come away with a greater appreciation for the technicalities of image compression and how your eyes work while also understanding how you can spot and reduce JPEG artifacts, among other things.

Susheel Chandradhas's picture

Susheel Chandradhas is a professional photographer and filmmaker based out of Chennai, India. He has a background in advertising and graphic design.

Log in or register to post comments
3 Comments

My brain hurts.

Much more complex than I thought.

JPEG tosses 90% or more of the capturable/captured source data, and we still love 'em!
.