Skip to main content

Deciphering yet another mystery parameter - camera picture quality

I don't know much about photography and much less about image formats. The little I know comes mostly from web development. There I use images as semi-static assets, and I know how to work around the potential problems. 


That experience was of limited use when I worked on end-user-created images on a React Native app I'm working on. I needed a crash course into JPEGs. 


We use react-native-camera to take photos. Snapping one has an elusive option called photo quality. It is an integer in a range between zero and one, and in the example code, it was set to 0.5. Zero point five. What does that mean? Is that good? 0.5 does not feel that good. To me, a modern iPhone is more than 0.5? The default value is 1 though, is that better? I wanted to find out what turning that mystery knob entails.


Spoiler alert: the JPEG quality is clearly explained in the JPEG Wikipedia article, but obviously, I needed to jump through some hoops to understand that. I want to go through some generic thoughts on how to approach these kinds of voodoo configuration parameters. Perhaps it is a suitable default, but who knows, right? Taking photos is a core feature, so I want to set the quality properly.


Initially, without knowing anything about JPEG compression and quantization tables, numerous things were puzzling:


  • As said, the default is 1, but the example code sets it to 0.5
  • Quality 0.5 intuitively seems low to me? Is it something that was added to the example years ago and is now outdated? Is that a good-for-all or a not-worst-case-for-all type of default?
  • What happens if I change that. Is that platform-specific? Will it affect performance? Indeed it will affect the file sizes
  • If I set it too low, the photos look terrible, but we save on our cloud costs. If I set it too high, the photos look great but are slow to load, and we need to pay more to the cloud provider. 
  • Do I need to do some A/B testing with the mystery knob?
  • How can I explain to the next developer that the mystery knob should be set to a specific value?


Figuring out that 0.5 is indeed a perfectly acceptable value for the app was awkward. It involved watching a YouTube video about JPEG, Intense googling about quantization tables, glancing through a whitepaper, reading some Android documentation, extracting JPEG quantization tables from a couple of test images, and finally, reading the Wikipedia article, which would have been the obvious step to start with.


Although the example code works for us, I think the default is an unfortunate choice, and the example code works only for applications not requiring a decent image quality. I'll settle that my lumpish exercise was not in vain. Once again, a lesson in the dangers of copy-pasting code from GitHub (although it's fine for 90% of the time 😉). 


PS: If you want to produce a "good enough" JPEG, use a quality value between 50-100. Q=50, which is the standard, is considered nowadays low.


djpeg can be used to extract the quantization tables. The above one is for the color, the below for luminance.


djpeg -verbose -verbose -verbose image.jpg > /dev/null 

This is the "standard" compression (Q=50)

I pulled a random image from Instagram and looks like it had less compression which is expected from applications where the images are a core feature

I assume IG does something clever, but this looks something like Q=75

Labels:

Comments

Post a Comment

Popular posts from this blog

RocksDB data recovery

I recently needed to do some maintenance on a RocksDB key-value store. The task was simple enough, just delete some keys as the db served as a cache and did not contain any permanent data. I used the RocksDB cli administration tool ldb to erase the keys. After running a key scan with it, I got this error Failed: Corruption: Snappy not supported or corrupted Snappy compressed block contents So a damaged database. Fortunately, there's a tool to fix it, and after running it, I had access to the db via the admin tool. All the data was lost though. Adding and removing keys worked fine but all the old keys were gone. It turned out that the corrupted data was all the data there was. The recovery tool made a backup folder, and I recovered the data by taking the files from the backup folder and manually changing the CURRENT file to point to the old MANIFEST file which is apparently how RocksDB knows which sst (table) files to use. I could not access the data with the admin tool, ...
Post a Comment
Read more

I'm not a passionate developer

A family friend of mine is an airlane pilot. A dream job for most, right? As a child, I certainly thought so. Now that I can have grown-up talks with him, I have discovered a more accurate description of his profession. He says that the truth about the job is that it is boring. To me, that is not that surprising. Airplanes are cool and all, but when you are in the middle of the Atlantic sitting next to the colleague you have been talking to past five years, how stimulating can that be? When he says the job is boring, it is not a bad kind of boring. It is a very specific boring. The "boring" you would want as a passenger. Uneventful.  Yet, he loves his job. According to him, an experienced pilot is most pleased when each and every tiny thing in the flight plan - goes according to plan. Passengers in the cabin of an expert pilot sit in the comfort of not even noticing who is flying. As someone employed in a field where being boring is not exactly in high demand, this sounds pro...
Post a Comment
Read more

PydanticAI + evals + LiteLLM pipeline

I gave a tech talk at a Python meetup titled "Overengineering an LLM pipeline". It's based on my experiences of building production-grade stuff with LLMs I'm not sure how overengineered it actually turned out. Experimental would be a better term as it is using PydanticAI graphs library, which is in its very early stages as of writing this, although arguably already better than some of the pipeline libraries. Anyway, here is a link to it. It is a CLI poker app where you play one hand against an LLM. The LLM (theoretically) gets better with a self-correcting mechanism based on the evaluation score from another LLM. It uses the annotated past games as an additional context to potentially improve its decision-making. https://github.com/juho-y/archipylago-poker
Post a Comment
Read more