Maximum PNG size according to resolution

1
votes

Is there a way to calculate the Maximum size that could take any image compressed with PNG ?

I need to know, that (for example) a PNG of a resolution of 350x350 (px), can't be larger than "X" KB. (and for a constant quality compression, like 90)

the "X" value is the one I'm looking for. Or in math expression 

350px *  350px * (90q) < X KB

I'm not quite familiar with the PNG compression algorithm, but there is probably a max value for a specific resolution ?

P.S. : the PNG has no alpha is this case.

1
compressionpng
Related, probably dup: stackoverflow.com/questions/22502529/… – leonbloy
You are right ! I didn't have the correct Tags when searching for my issue. – TheSquad
Do not mix up 'resolution' and 'size'. Resolution is an absolute value: the number of pixels per (absolute) measurement unit. A typical unit is "pixels per inch", or 'ppi'. What you are asking about is size, in pixels only. – Jongware

1 Answers

4
votes

In the maximum case, the data is incompressible (for example, if the image size is 1x1, or if the image is larger but contains random incompressible data). So the maximum size is

8 // PNG signature bytes
+ 25 // IHDR chunk
+ 12 // IDAT chunk (assuming only one IDAT chunk)
+ height //pixels
  * ( 1 // filter byte for each row
      + (width // pixels
         * 3 // Red, blue, green color samples
         * 2 // 16 bits per color sample
        )
    )
+ 6 // zlib compression overhead
+ 2 // deflate overhead
+ 12 // IEND chunk

Compression "quality" doesn't enter into this. Most applications will probably separate the IDAT into smaller chunks, typically 8kbytes each, so in the case of a 350x350 image there would be 44 IDAT chunks, so add 43*12 for IDAT chunk overhead.

As a check, a 1x1 16-bit RGB image can be written as a 72-byte PNG, and a 1x1 8-bit grayscale image is 67 bytes. If the image is interlaced or has any ancillary chunks, or has an alpha channel, it will naturally be bigger.