Can someone explain Lightroom file size exports to me?

[smmokan]

I am putting together some files from my recent trip to print out to canvas, and I've noticed something funky with the file sizes when I export from LR3.

The first file is my Monument Valley sunrise shot, seen below. The image file is about 75-80% of the original size- meaning I cropped out about 15-20% of it- and when I export the file to JPG it's only 4.10MB despite having dimensions of 5413 x 2705.



The second file is the House on Fire shot, which is approximately the same crop factor, about 80%. Its measurements are 3071 x 4606 but when I export it, the file size is 17.0MB.



What's up? Even though they're approximately the same dimensions pixel-wise, they export to significantly different file sizes. Is it because the information in each pixel is different due to more light being captured in the second shot? I didn't edit them differently, and I exported them with the same settings. Maybe my brain is still on vacation.

[systemoverblow'd]

I think it's the detail. jpeg compression is based on near pixels. In the top picture tou have many planes of very similar pixels. This is helpful for compression algorithms. The bottom one, not so much. Or that's what I assume is happening. Happens that way for video.

[systemoverblow'd]

Let me explain how a video codec works and I assume jpeg is similar. It will sample a single pixel and write the RGB value as 264,10,78 or something. Then it will go a few pixels over and write the data of that pixel as say 38,100,120. Depending on how fine quality the codec is, the closer those pixels are that are smapled. Heavy compression samples further pixels. What the codec then does is interpolates the pixels between the sampled ones as a a gradient range. How pixel #1 R value gets to pixel #4 R value by assigning R values across those pixels it didn't sample.

Does that make sense? You have massive chunks of the same color so there isn't a lot of information to deduce in color changes, therefor it doesn't need to take up as much space. Again, at least that is how video frame compression works so I assume in theory it is very similar to still jpeg compression. It's alos why heavily compressed video like h.264 is small but very processor intensive. The computer has to "think" about how to recreate each frame while a lossless codec file is gigantic in file size because every pixel value will be written into the file.

[systemoverblow'd]

These 2 identical size jpegs that I just created are 25kb and 93kb respectively.

Just black and white and almost 4 times the size.

[Bronic]

I was curious about the same thing a while back while noticing various file sizes making panoramas. Same focal length and res. but very different sizes. I did some research and found the same stuff system was taking about

[smmokan]

Ok, that's kind of what I figured, but I was surprised it was such a big difference.

[Hugh Conway]

jpeg is also lossy compression; if you don't think this matters there's jpeg porn out there that's been done in thousands of times since the 90s that sucks ass now.

[Altmanator]

Were the ISOs different? When you're shooting at a higher ISO your file size will be different.

Hmm, now that I think about it I have noticed that with RAW files but I'm not sure I've checked with JPGs.

Now I'm curious, are ISOs the same with both photos?

[systemoverblow'd]

Higher iso's will cause more grain or noise. That will make big differences in color between pixels therefor higher file size.

[smmokan]

The sunset was taken at ISO 400, the House on Fire was ISO 500. Not much difference there.

[Chainsaw_Willie]

What Scott said is basically what's going on. The most basic interpretation is that when you have large areas of pixels with the same color values lined up together it's easier for the compression algorithm to do it's job. When a picture contains lots of fine detail and pixel-to-pixel variations in color it's harder for it to compress the image.

Higher iso images tend to have more noise, and thus more variation in color from one pixel to the next. Images with a lot of sharpening applied to them will also end up with larger file sizes - all else being equal, because sharpening induces more variation in pixel-to-pixel color.

Your sunset pic has large areas of pixels with the same color values all lined up together. The compression algorithm basically looks at the image line-by-line and says "the next X pixels are color Y". That bit of information can be stored in a byte or two. When the pixel color information changes it stores another "next X pixel(s) are color Y" value, and so on. In a pic like your sunset pic you might have 500 pixels (for example) in a row that are all the same color. Each individual pixel's information doesn't have to be stored, just a single value for how many pixels in the sequence and another value for what color.

In your second pic of the Anasazi house there are not these large blocks of same-colored pixels so the compression algorithm has to store a lot more individual values. Instead of 500 adjacent pixels all having the same color value and being stored in a small package, that file is going to have a lot more instances of only 2 or 5 or 10 pixels adjacent to each other having the same value.

I'm sure the compression algorithm is actually much more complex than that, but in a nutshell that's what it's doing and why a picture with large blocks of solid colors will be much smaller than a pic with lots of fine detail and variability from one pixel to the next.

[MarcusBrody]

Also just to make sure: you had the JPEG compression quality slider set at the same level for both, right? Even when I'm exporting at the same nominal pixel size, I've found that my jpeg file size is very sensitive to where I set it, particularly at the top end of the of the quality scale. I just did a random sample (16mp) photo and the file at 100% was 2.5 times bigger than the exact same image at 85%.

[smmokan]

Yep, confirmed... I export everything at 100% and never touch the quality slider.