Reduce resolution of image in 2d array

[Isaac-Leonard]

I've got pixel data in a 2 dimensional array and I want to reduce the resolution by splitting into chunks and running interpolation on it such as using the .mean() method.
So something along the lines of

let img = img.chunk_2d(x_scale, y_scale).map(|chunk|c hunk.mean());

The chunk_2d method doesn't exist of course, is there a similar method that would do what I want or how would I go about doing this?
I've looked at the .into_shap() method but it doesn't seem to do what I want exactly.

[Isaac-Leonard]

Also the .chunks_exact() method doesn't seem to give a good way to resolve back into a 2d array and only gives one dimension to chunk the data by so can't have different x and y scales

[nilgoyette]

Would this work?

let mean = Zip::from(img.exact_chunks((x_scale, y_scale))).map_collect(|chunk| chunk.mean());

exact_chunks is a producer, so it can be used with Zip. We usually zip over more than one array, but this is not mandatory. Zip is a wonderful tool! I suggest you to read it's doc.

[Isaac-Leonard]

It looks like that works, thank you
Only tangentially related but would you know how I could make this so I'm not iterating over references to the pixels in:

    for row in data.rows() {
        for (i, pixel) in row.into_iter().enumerate() {}
    }

I'd rather iterate over the raw values Or do you think dereferencing wouldn't matter performance wise?

[nilgoyette]

I can't remember where this is written, but you should use iter/into_iter as less as possible because it is slow. For example, what you wrote can be expressed as

Zip::indexed(&data).for_each(|idx, &pixel| {
    if idx == (0, 0) { wathever }
})

If you're starting with ndarray, I strongly suggest you to read the quickstart and official doc.

[Isaac-Leonard]

Awesome, thank you so much and yeah I'll do that now.
I kinda just jumped straight in lol