Author Topic: 4.6 problems  (Read 12663 times)

Offline Kirk Baker

  • Senior Software Engineer
  • Camera Bits Staff
  • Superhero Member
  • *****
  • Posts: 25056
    • View Profile
    • Camera Bits, Inc.
Re: 4.6 problems
« Reply #15 on: March 29, 2009, 06:18:58 PM »
Russ,

OK, it's a little better in the 3-28 version, the estimate was 44KB and the result was 152KB.

That difference could easily be explained by EXIF, IPTC, XMP, and ICC profile data sizes not being part of the size calculation.

-Kirk

Offline Boatguy

  • Member
  • **
  • Posts: 69
    • View Profile
Re: 4.6 problems
« Reply #16 on: March 30, 2009, 04:27:24 PM »
OK.

However, it may be worth noting that the same imgae, saved to the same size in 4.5.4 had a predicted size of 43KB and a final result of 108KB.  Larger than the prediction, but closer than 4.6.1 which resulted in 152KB.

I assume that the EXIF, IPTC, XMP, and ICC profile data is a known constant, certainly known for a given source image.  It seems like it could be folded into the predicted size with a great deal of certainty.

For my usage in producing images for a web site the resulting size is important.  If you're saying it's as good as it can be, then I have no choice but to accept that, but it's disappointing that the accuracy declined from 4.5.4 to 4.6.1.

Offline Kirk Baker

  • Senior Software Engineer
  • Camera Bits Staff
  • Superhero Member
  • *****
  • Posts: 25056
    • View Profile
    • Camera Bits, Inc.
Re: 4.6 problems
« Reply #17 on: March 30, 2009, 04:33:04 PM »
OK.

However, it may be worth noting that the same imgae, saved to the same size in 4.5.4 had a predicted size of 43KB and a final result of 108KB.  Larger than the prediction, but closer than 4.6.1 which resulted in 152KB.

I assume that the EXIF, IPTC, XMP, and ICC profile data is a known constant, certainly known for a given source image.  It seems like it could be folded into the predicted size with a great deal of certainty.

For my usage in producing images for a web site the resulting size is important.  If you're saying it's as good as it can be, then I have no choice but to accept that, but it's disappointing that the accuracy declined from 4.5.4 to 4.6.1.

There are now 100 quality settings versus about 10 for the earlier version, mostly on the low end is where the additional quality values are found.

The size estimation is for the first photo in the selection so if it had a crop and the others in the selection did not then that could produce some confusion.

Are you only working with a single image when you're doing these tests?

Also, for the 152 KB file produced in 4.6.1, what was its size estimate?  And, if you want me to look into it further, please send me a sample image and I'll see how much non-jpeg data there is in the file. Please click on my name to the left of this message, then click on the 'personal message' link.  I will respond with upload instructions.

-Kirk

Offline Boatguy

  • Member
  • **
  • Posts: 69
    • View Profile
Re: 4.6 problems
« Reply #18 on: March 30, 2009, 05:43:16 PM »
Yes, the test was for a single image.

As reported earlier "OK, it's a little better in the 3-28 version, the estimate was 44KB and the result was 152KB."

I'll send you the image.

Thanks,

Russ