I am resizing an image from 3000x810 to something smaller 900x243 but this makes the image blurry. I am curious why this is happening. I understand why making an image larger results in a blurry/pixelated look but not sure why this is happens when getting smaller.
This question needs some examples which needs some time to be prepared. I will complete it later.
Very nice question.
I would say that it depends on several factors which I will list prior to the explanations.
1. The resampling method. Most of them make an average of the original pixels, except one that only uses one row of the original pixels.
2. The original data. How blurry or sharp the original image or zone is.
3. The scaling factor. Basically either using round numbers (200%) or something that needs to make some averaging calculations (150%)
4. The viewing size. You can see an image either at a specific screen size, for example at the full height of the monitor, or relative to the dimensions of the image in pixels, for example at 100%.
Let me explain some hypothetical cases.
Imagine you have one image that you will use as a wallpaper on your computer, and you have a FULL HD, 1920x1080 px.
If your image is at the native resolution it will display 1 image pixel on every screen pixel. So far so good.
If your image is smaller, the pixels will try to invent new in-between information, and a typical way to do it is averaging the data (blurry) or duplicating the data (nearest neighbor)
Depending on the "sharpness" of the original image one or the other method will be better. A normal photograph vs a vintage computer graphic. In the second case, a pixel art will retain sharpness even if upscaled, especially if the upscaling is at perfect multiple times (200%-300%-400%)
When you downscale the image, for example squeezing a 4K wallpaper on your FullHD monitor, the same way you could need to average the pixels to fit them in the smaller space.
This means that any averaging actually makes an image blurrier.
But if the original image is actually blurry, like an out of focus image, resampling down with the nearest neighbor method, dropping one row of useless duplicated data, will make the image sharper per group of pixels.
Here is the catch, the sharpness of an image is dependent on the neighboring pixels if these adjacent pixels have more contrast the image will look sharper.
But there is also another element to be considered.
5. Perceived result.
If your image had some little text, and with the downsampling, you can no longer read it you could perceive it as "blurrier" because you actually loosed information.