How to improve quality of a low res jpg please

Messages
22
Edit My Images
Yes
Hello

I have a photo of low res that I want to improve, good enough to have a large print for my wall. I understand that adding pixels lowers that resolution similar to removing them.

Seems strange, but what do I do to improve the resolution please?

Regards
 
Have not you still got the original file?

When you say low res, just what have you got in pixel dimensions and what file size is it? Oh, how big a print are you looking to produce?
 
You cannot increase resolution after the shutter has fired or after resolution has been reduced by processing. Adding pixels does not alter resolution and will stop you seeing individual pixels in a large print - but it will still be low resolution.
 
There's a limit to what can be done - you're handicapped so must moderate your requirement. Experiment. It might be a jpg? In which case it'll be pixellated and show jagged edges etc if you enlarge it - try it at different sizes and view the results at 100% on screen. Ok - go back to the original and apply a bit of Gaussian blur. Experiment with the amount, but it won't be much. Then enlarge it modestly, and view the result - it should look better than the unblurred version, but it'll be on the fuzzy side. Apply a teeny bit of sharpening and see what happens. And that's going to be about as good as it gets.

If it's a tif, that's a much better starting point.
 
This often gets slightly better results than photoshop's normal upscaling:

https://www.on1.com/products/resize/


There's a limit to what can be done - you're handicapped so must moderate your requirement. Experiment. It might be a jpg? In which case it'll be pixellated and show jagged edges etc if you enlarge it - try it at different sizes and view the results at 100% on screen. Ok - go back to the original and apply a bit of Gaussian blur. Experiment with the amount, but it won't be much. Then enlarge it modestly, and view the result - it should look better than the unblurred version, but it'll be on the fuzzy side. Apply a teeny bit of sharpening and see what happens. And that's going to be about as good as it gets.

If it's a tif, that's a much better starting point.

Adding a little noise after sharpening can help too.
 
The problem with magnifying an image is that it will magnify some of the unpleasant details of the original image-making process to the point where they become annoyingly visible. For example if you magnify a newspaper photograph you'll start to see all the dots it's made of. If you magnify a digital image you'll start to see the square pixels it's made of. If you magnify an old monochrome film print you'll start to see the individual grains of silver oxide. Often however the important details of the original image, such as the brush strokes of a painting, are much larger than these annoying artefacts. In which case it may be possible to blur out these artefacts and then sharpen back the edges of the brush strokes (or whatever) to produce a much more acceptable magnified image.

That's what the best enlarging systems try to do, with varying degrees of success. The amount of success depends critically on the nature of the original image. Sometimes a combination of methods with careful manual intervention produces the best results.
 
Topaz labs have just released a new product that makes pretty big claims... https://topazlabs.com/ai-gigapixel/

This. Let AI hallucinate its own thing and pray that it gets it "right". There is no guarantee at all...
I imagine there are obvious things like lines and other geometric shapes in the objects, the usual patterns like foliage and so on. The more disorderly fine detail the harder it gets. It AI is trained well it may recognise some artefacts and attempt to remove them. If the file is both noisy and defocused the chances are getting very slim.
 
This. Let AI hallucinate its own thing and pray that it gets it "right". There is no guarantee at all...
I imagine there are obvious things like lines and other geometric shapes in the objects, the usual patterns like foliage and so on. The more disorderly fine detail the harder it gets. It AI is trained well it may recognise some artefacts and attempt to remove them. If the file is both noisy and defocused the chances are getting very slim.

From what I have seen AI doesn't attempt to get a clear image from a poor image (although the fact that you are working with a larger image can help with PP) just to enlarge what is already there.

One odd thing about it is that one reviewer has produced an image from the program where it altered the slope of a roof!

https://www.dpreview.com/forums/post/61509742

But it still does a very good job of producing very large, artefact free images.
 
Last edited:
Hello

I have a photo of low res that I want to improve, good enough to have a large print for my wall. I understand that adding pixels lowers that resolution similar to removing them.

Seems strange, but what do I do to improve the resolution please?

Regards


Well, I've not tried it, but maybe others could say if it is good or bad idea, how about this...

Print out the image at its current size on best glossy paper you can get hold of. Don't try expand the image, like fit to paper, just leave it as it is in its original size. Set the printer's settings to maximum possible, in other words, depending on printer make and model, find highest quality setting, like Best rather than Fast, Print it out.

Stick the printed out image on wall, then set up a camera on tripod (or any other support you can use) at best possible distance from the printed out copy. Make sure it fills up the viewfinder screen. Don't use direct flash as it may reflect off the glossy paper. Use camera on lowest ISO setting possible (like 100) to avoid noise grain, but at the camera's maximum resolution and highest quality setting (For example: With a Nikon, it's set to Fine).

Take a photo. Hopefully you have a second photograph at the highest resolution of the original image at the lowest resolution.

Just don't try to resize the new photo to a smaller resolution or compress it too much to save desk space, otherwise, next time you open it and attempt to upscale it, you'll be back to having blocky pixels, leave the new copy at the highest resolution.

No idea if it's of any help, but I'm guessing it's better than nothing.
 
Well, I've not tried it, but maybe others could say if it is good or bad idea, how about this...

Print out the image at its current size on best glossy paper you can get hold of. Don't try expand the image, like fit to paper, just leave it as it is in its original size. Set the printer's settings to maximum possible, in other words, depending on printer make and model, find highest quality setting, like Best rather than Fast, Print it out.

Stick the printed out image on wall, then set up a camera on tripod (or any other support you can use) at best possible distance from the printed out copy. Make sure it fills up the viewfinder screen. Don't use direct flash as it may reflect off the glossy paper. Use camera on lowest ISO setting possible (like 100) to avoid noise grain, but at the camera's maximum resolution and highest quality setting (For example: With a Nikon, it's set to Fine).

Take a photo. Hopefully you have a second photograph at the highest resolution of the original image at the lowest resolution.

Just don't try to resize the new photo to a smaller resolution or compress it too much to save desk space, otherwise, next time you open it and attempt to upscale it, you'll be back to having blocky pixels, leave the new copy at the highest resolution.

No idea if it's of any help, but I'm guessing it's better than nothing.

Why would you think it would do anything positive other than adding more noise, distortion, softness and possibly glare?

I am strongly of an opinion that you can't polish a turd (no pun intended) but it may be possible to pretty much repaint it by hand or with help of software...
 
From what I have seen AI doesn't attempt to get a clear image from a poor image (although the fact that you are working with a larger image can help with PP) just to enlarge what is already there.

One odd thing about it is that one reviewer has produced an image from the program where it altered the slope of a roof!

https://www.dpreview.com/forums/post/61509742

But it still does a very good job of producing very large, artefact free images.

Topaz ran a webinar last week demonstrating AI Gigapixel and the results looked extremely impressive.
The recording of the webinar is here:
View: https://www.youtube.com/watch?v=JAL2q_Lhlw8

The famous expression "garbage in = garbage out" applies here, with the final result very much dependent on the quality of the original.
One thing it does NOT do is take a rubbish image and transform it into something wonderful.

For those who wish to try the program, so they know what they are talking about, you may download a 30 day free trial.
https://topazlabs.com/ai-gigapixel/
 
Firstly thank you everybody.

Hello Box Brownie the sizes are 1600 x 1200 and the print size I would like is 30" x 22.5".
Hello DG. See above, thanks.
John.M. Yes this is my question. I did give a picture of fairly low res to a firm who did produce a large paper print, I would like to know how to do it.
Droj and juggler, I will try that thanks.
Major Easy and Long lens. A good practical idea, I shall certainly try that. Thank you.

Thank to everyone else, I wonder how this firm did it?
Best
 
In the past I have had prints done with a request to the lab to upscale them but nothing like your >400% increase (@300ppi). The company I used had sophisticated software (I understand called a RIP ~ I looked into such a RIP at the time and found one selling licensed RIP software for more than £1000, so very much the pro lab usage) and said that they can sometimes depending on the type of image get quite large increases......no idea how large though.

You gave the size but did not say why you don't have a larger sized original???
 
Last edited:
1600 x 1200 and the print size I would like is 30" x 22.5"

That's a bit of a stretch tbh - but you certainly don't need 300 dpi at that size

When I used to buy bigger canvases 100 dpi was more than enough for a 3ft wide image (a bit bigger than yours), so I'd take a look at any software to double the longest side and just buy a test print - a cheap one from such as DSCL - to see what you think

Canvas prints hide low res better than a normal print

Oh and the reason you don't need 300 dpi is because you're going to hang it and view it from several feet away, not stick your nose on it looking for 'quality' - the farther away you view anything the lower the actual resolution that's needed; so get that test print and if its acceptable from a few feet away you're sorted to get a better quality one

Dave
 
Can we see the image? Might help to decide what can be done with it.
 
I wonder what software they are using here....

 
That's a bit of a stretch tbh - but you certainly don't need 300 dpi at that size

When I used to buy bigger canvases 100 dpi was more than enough for a 3ft wide image (a bit bigger than yours), so I'd take a look at any software to double the longest side and just buy a test print - a cheap one from such as DSCL - to see what you think

Canvas prints hide low res better than a normal print

Oh and the reason you don't need 300 dpi is because you're going to hang it and view it from several feet away, not stick your nose on it looking for 'quality' - the farther away you view anything the lower the actual resolution that's needed; so get that test print and if its acceptable from a few feet away you're sorted to get a better quality one

Dave
^^^^ This... but change dpi to ppi please ;) :exit:
 
^^^^ This... but change dpi to ppi please ;) :exit:

Except that dpi is the printer setting to tell the printer what resolution to print at, so DPI is correct here lol - I hope this helps from Wiki...


"Dots per inch (DPI, or dpi)[1] is a measure of spatial printing or video or image scanner dot density, in particular the number of individual dots that can be placed in a line within the span of 1 inch (2.54 cm).

Monitors do not have dots, but do have pixels; the closely related concept for monitors and images is pixels per inch or PPI. Many resources, including the Android developer guide, use the terms DPI and PPI interchangeably."


Dave
 
Except that dpi is the printer setting to tell the printer what resolution to print at, so DPI is correct here lol
Oh no it's not... dpi does in fact refer to printing but refers to the number of dots a printer lays down per inch.

An image is made up of pixels not dots, a pixel can consist of many printed dots. The two terms have become interchangeable through mis-use.

I am sorry Dave you have that wrong :p
 
Oh no it's not... dpi does in fact refer to printing but refers to the number of dots a printer lays down per inch.

An image is made up of pixels not dots, a pixel can consist of many printed dots. The two terms have become interchangeable through mis-use.

I am sorry Dave you have that wrong :p

I'll have to look into that more before conceding lol - but yes they are now generally considered the same thing, so its a pointless argument

That I still want to win :D

Anyway - glad we agree on his course of action :)

Dave
 
@PhilH04 - see this from Loxley - top printer in the UK...

Images should ideally be sized at the required print dimensions at 300ppi (pixels per inch) and saved at Photoshop Level 10 JPEG as Baseline (Standard). However, lower resolutions can still result in a high quality print*.

So there then !!!

b****x :D

You win

Dave
 
^^^^ This... but change dpi to ppi please ;) :exit:

Except that dpi is the printer setting to tell the printer what resolution to print at, so DPI is correct here lol - I hope this helps from Wiki...


"Dots per inch (DPI, or dpi)[1] is a measure of spatial printing or video or image scanner dot density, in particular the number of individual dots that can be placed in a line within the span of 1 inch (2.54 cm).

Monitors do not have dots, but do have pixels; the closely related concept for monitors and images is pixels per inch or PPI. Many resources, including the Android developer guide, use the terms DPI and PPI interchangeably."


Dave

Oh no it's not... dpi does in fact refer to printing but refers to the number of dots a printer lays down per inch.

An image is made up of pixels not dots, a pixel can consist of many printed dots. The two terms have become interchangeable through mis-use.

I am sorry Dave you have that wrong :p

@PhilH04 - see this from Loxley - top printer in the UK...

Images should ideally be sized at the required print dimensions at 300ppi (pixels per inch) and saved at Photoshop Level 10 JPEG as Baseline (Standard). However, lower resolutions can still result in a high quality print*.

So there then !!!

b****x :D

You win

Dave

:cough: I did say "ppi" in my post #19 but Dave decided to talk about "dpi" in post #20.........just saying...;)
 
Oh no it's not... dpi does in fact refer to printing but refers to the number of dots a printer lays down per inch.

An image is made up of pixels not dots, a pixel can consist of many printed dots. The two terms have become interchangeable through mis-use.

I am sorry Dave you have that wrong :p
My printer prints just over 14,000 dots per inch. If the image has 300 pixels per inch, that is nearly 500 dots per pixel.
 
Well I have started an interesting fight.

I have a lot of reading to do, your comments have helped a lot. Thank you all.

Best regards to you.

Alan
 
the only progam i know that is capable of inceasing size without losing as much quality is adobe illustratrator

But surely a graphics resizer is not that efficient when resizing jpegs. I.e. AFAIK the difference between vector and raster images!
 
vector images can be scaled without losing quality, so if an image is converted to vector , it shouldnt lose any more quality, or at least lest than a non vector processor
 
a jpeg can be resized in a graphic resizer by 150% then the quality drops significantly, vector the only limit is the amount of memory you have.
an example i once took a 35 mm slide scan to 1m pixels before i ran out of memory and the quality was still there
 
vector images can be scaled without losing quality, so if an image is converted to vector , it shouldnt lose any more quality, or at least lest than a non vector processor

a jpeg can be resized in a graphic resizer by 150% then the quality drops significantly, vector the only limit is the amount of memory you have.
an example i once took a 35 mm slide scan to 1m pixels before i ran out of memory and the quality was still there

I live & learn but the 150% might not meet the OP's needs, if enlarged as a vector image what happens when you convert back a jpeg???
 
Back
Top