Megapixels vs PPI, I don't get it.

I'm confused why an exported file of 13.3MB with dimensions of (4752 X 3168) = 99PPI and a smaller file 11.5MB has dimensions that are larger at 5573 X 3715 = 116PPI. This is regarding a print size of 32" X 48" for both images.

I thought the higher the megapixels, the higher the resolution / size of the image, larger the print can be. Can someone shed some light on this for me? I am looking to print a few canvas and it looks like the 99PPI just misses the recommeded 100-150PPI.

Thanks for any help.

For file size, the PPI is not relevant. PPI only matters when you go to actually put them image onto paper. How big is an inch on your hard drive?

As for the file size in megabytes: Are they the same image? If not, you can't really compare them. It's black magic... I mean, a complex jpeg compression algorithm that's hard for mere mortals to predict. Different image contents will handle the compression differently.

Now, if it's two exports of the same image, I'm running out of ideas. If they're the same image, exported at two different pixel dimensions, and with the same level of jpeg compression, then the one that's larger in pixel dimension should be larger in file size, I would think.

...although, on second thought: If it's resized upon export, then it's no longer the same image. Every pixel is different and new, and will be handled differently by the jpeg compression algorithm.

divide pixels by inches and you get pixels per inch.

Second file had 5573 pixels and 48" on the long side so 5573/48 = 116 PPI.

Megabytes are just a measure of storage space on your disk or mem card, they are irrelevant to the equation.


And honestly, if 100 is in range, 99 is fine. You won't be able to tell the diff. Or just print it 46" instead of 48 (or next smaller size) if you are worried.

I just made a small section into two different size prints...one at 153 pixels per inch (306 pixels across 2" print area) and the other at 55 pixels per inch (306 pixels across 5.5" area)...and then scanned both images at 300 dpi to compare...

You are confusing megabytes with megapixels. Unless you are talking about uncompressed tifs there is no direct correlation between them. Since you are apparently talking about jpg files, which are always compressed, comparing MBs means nothing. Not only might they be compressed different amounts, the same compression setting produces different sized files depending on the amount of detail/noise in the image.

5573 X 3715 = 20.7 MegaPixels
4752 X 3168 = 15.0 MegaPixels

Just to reiterate what others have said. For JPEG image files the size of the image file on the hard drive has minimal relationship to the number of pixels in the image.

In order to demonstrate this I produced a new file in PSCS5 3072 pixels wide by 2048 pixels tall and filled with the default background colour, white. It could have been any other colour, just as long as it was perfectly uniform. The reason that I picked those exact pixel dimensions is that for an image that uses 8 bits of data for each colour channel it requires EXACTLY 18 Megabytes of memory to hold the image.

So I then saved the image as an uncompressed TIFF file, this file according to Windows Explorer occupies 18456 Kb. the odd differences is from the extra data that the file uses to hold things like the EXIF data and colour space information that is needed, but doesn't actually make up the image. I then saved the image with four different levels of JPEG compression: 0, 6, 9 and 12, from lowest to highest quality. I picked these levels as the complicated JPEG compression algorithm is not linear, so the biggest differences in file size/quality setting come between levels 6 and 12. so for this plain image with ZERO detail the JPEG file sizes are Q12: 421 Kb, Q9: 200 Kb, Q6: 127 Kb and Q0: 86 Kb. This is simplest option for the JPEG algorithm as there is no detail, so the difference between Q0 and Q12 is not observable when you open the actual image.

Now to give the image lots of detail I then went and added noise to the image. I used the Add Noise filter, with a setting of 100% and a Gaussian distribution, this was colour noise. This makes just about every pixel different from the 8 others adjacent to it. This is the worst possible case for JPEG compression to deal with. So I saved the same set of five files. The TIFF file came out at 18462 Kb this time, the extra data stored is always slightly different for each file, but as the image data is stored with no compression it is always the same size. The JPEG file sizes though are very different. Q12: 21035 Kb, Almost 3 Mb bigger than the TIFF file. Q9: 13537 Kb, Q6: 5593 Kb and Q0: 4632 Kb. Because it is very hard for the JPEG, or any other system of compression to deal with such highly detailed images is why these file sizes stay so large. It did surprise me that the Q12 file was so much bigger than the uncompressed TIFF file.

Finally I added a Gradient to the noise image so that it went from plain white at the top of the image to the noise pattern at the bottom. This was to represent a half way situation, that although controlled is a bit more like a normal photographic image. The results from this test were roughly in line with my expectations. The TIFF was this time 18466 Kb, JPEG Q12: 13646 Kb, Q9: 7961 Kb, Q6: 3135 Kb and Q0: 2146 Kb.

So as you see the number of pixels: 6291456 in this case, has no relationship to the different compressed JPEG file sizes achieved, that was all down to the content of the image, and the quality setting used. The uncompressed TIFF file though, within the differences of the extra bits needed for things like EXIF stayed constant at 18 Mb, as TIFF file sizes (for images with a single layer) are wholly dependent on the number of pixels, not the content. This compressed file size business is an issue that many have a job getting a handle on.

I have added the files to my Dropbox here.

As for your PPI issue. The more PIXELS you have the higher the PPI will be for any PARTICULAR print size. If you spread those pixels over a larger area the PPI number will go down. You will not see any difference between 99 and 100 PPI on any type of print, and canvas is a naturally low quality medium, compared to say a glossy print, hence the much lower requirements.

Alan

I was totally confused and the explanations have helped a lot. I really appreciate the feedback

Your digital image is basically made up of lots of little square tiles all arranged neatly in rows.

Megapixels are how many tiles are in the image. The 7D, for example, produces images that have 5184 tiles going from side to side and 3456 tiles from top to bottom. This means that there are 17,915,904 tiles in total, or 17.9 million tiles. Megapixels are measured in millions of pixels, so the 7D has 17.9 megapixels.

PPI is pixels per inch only comes up when you are printing or displaying the image. It refers to how big each tile is. If each tile is one inch across, then the image is being displayed at 1 PPI. If you looked at a picture like this closely, then it would appear very blocky. But the whole image would be 432 feet across. People would be looking at a picture this size from quite a long way away, so they won't really notice that the pixels are so big. If you shrink the size of each tile down, then it won't look so blocky, but you won't have enough tiles to make the picture so big. Generally, when you print a photograph, you should have a PPI of about 200 - 300 ppi. This will be enough to make sure that people can't see the individual pixels in the image.