1

I know there are a few questions regarding reducing image size while minimizing image quality, and my question relates to it, but I'm more concern about what the initial image size should be to minimize as much quality loss as possible.

For example, let's say I want to downsample an image to a size of 500x500.

To minimize as much quality loss as possible, is it best to find the highest quality resolution image of the biggest size, i.e. 1000x1000 or 5000x5000, etc., or somewhere near the size I want to resized to, i.e. 600x600 or 700x700, etc.?

My thought process is that if the number of pixels to downsample to is minimized, then the image quality is minimized, i.e. if I downsample a high-quality image of 600x600 to 500x500, then there is much less image quality loss then downsampling from 1000x1000 to 500x500.

Is my thinking wrong or can I get better insight in this?

Thanks.

Improve this question
2
  • What kind of image are you resizing? Is it a photograph, or a graphic with fine detail that is being lost? What have you tried? What has failed/gone wrong? What software do you have access to?
    – Billy Kerr
    Dec 12, 2017 at 13:05
  • Are you asking for a tutorial?
    – user9447
    Dec 13, 2017 at 12:48

1 Answer 1

Reset to default
1

The best resolution to start an image of 500x500px is... 500x500px.

I have the feeling that you want to reinvent the wheel or something. But lets explore the matter a bit.

Any resampling or resizing of an image means it has some manipulation on the data. This not always means "loos of quality" but it implies manipulation of the information, the information carried by each pixel is somehow manipulated.

Depending on the type of image, a graph, a vector drawing, an heraldic shield or a modern simple logo, a photo of a gothic cathedral or a photo of a 70's furniture... it can react good or not so good to this change in information.

Another factor is your source device, if you have a camera of 24Mpx to take your picture, the best resolution would be... taking the original photo at 24Mpx.

The neverending recommendation. If you can make an image in vector format, do it, if you have a decent camera that takes nice pictures, use it.

A couple of specific recommendation I would make. If you are going to make a digital paint, do it about the same size of a 24Mpx photo, 6000x4000px. You will not have a slow monster to work with, but at the same time, you have a good size to resample to whatever size you want.

Of course, resampling it to 500x500px will need a crop and you will lose a lot of details, but you have your original with you, probably that 500x500px image is only a thumbnail.

There are times where the antialiasing system of the program where you are creating the image is not the best one. For example, a render engine of a 3D program can render some noise, or the aliasing of diagonal lines on the exported vector logo can look like a saw. In those cases go to the simplest formula. Make it twice as big, so generate a 1000x1000px image. Now you can play with the settings of your raster program, use a bicubic, sharpen it or whatever you want.

Improve this answer
1
  • I guess I should have been clearer in my post if I wasn't. To clarify, I don't have an image of 500x500. I do have an image of 1000x1000 and 600x600 and would like to downsample it to 500x500 to display in my app. I am using PhotoShop with the Bicubic Sharper (best for reduction) to downsample it and was asking about what the best initial size image to start off with to maintain the least minimal manipulation on the image data.
    – Pangu
    Dec 13, 2017 at 19:55

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.