Resolution ≠ Image Quality? What is the Principle of Image Compression at Constant Resolution?

Suppose we zoom in on an image and look at just eight pixels, it would look like this.

In a computer, this color block should be recorded like this (hypothetically):"Position 1: red, Position 2: orange-red, Position 3: pink, Position 4: Yunnan red, Position 5: green, Position 6: dark green, Position 7: light green, Position 8: medium green"

Right? Makes perfect sense.

Then I invented aPinecone Compression Algorithm, and after compressing this image, it turned into this.

So how can we record this image now?

"Positions 1-4: red, Positions 5-8: green"

Done.

Still eight pixels, but obviously, the amount of information needed to record is reduced, at the cost of losing color details.

Refining this algorithm logic just a tiny bit more makes it one of the most basic image compression algorithms.

Of course, modern image compression algorithms have become much more complex—this is just the most fundamental demonstration.

At this point, I was about to hit send, but looking at the image, I thought some friends might still not understand and feel like this is just reducing eight pixels to four, clearly lowering the resolution.

Let me give another example, still using the same algorithm.

This is before optimization

This is after optimization

After optimization, it can be recorded as "red, green×3, red×3, green"

Clearly, it’s still much more concise, and compared to simply lowering the resolution, the advantage is that it retains a tiny bit more detail.

In other words, compression algorithms that don’t reduce resolution try to preserve as much information as possible, performing better than simply lowering resolution.

Writing this, as I was about to hit send again, I thought since we’ve come this far, why not demonstrate with an actual image.

So I found an image—the photo below is of my desk, which I took yesterday.

Zooming in on the eye details of Hutao in the image above, the original looks like this. The file size at this point is 3.62MB.

Then I directly used Photoshop to compress it without reducing resolution, shrinking its size to just 313KB.

Now, looking at this detail again, the pixels have changed to this.

Everyone can take a closer look at the details. The number of pixels in both images is the same, but the colors in the second one clearly show a lot of smudging.

Other areas might not be as obvious, but right in the center of Hutao’s eye, you can see a square-shaped color block.

This is caused by the compression algorithm.