| The patent doesn't cover JPEG (Score:5, Interesting) by StevenMaurer on Thursday July 18, @02:57PM (#3910717) (User #115071 Info | http://slashdot.org/) |
| I have a combined ten patents issued and in process in this specific field, so I believe I can call myself an expert in this matter. The claims in this patent cover digital streams which tend to come in tuples, possibly with appended data. Something like this: (1,4) (1,3), (1,6), (4,6), (3,6), (9,6) It specifically claims the separation of these tuples into separate run-length encoded streams. In my example above, it might be: (3x1, 4, 3, 9) (4, 3, 4x6) There are some further claims about coding signs and amplitude, and some table lookup mechanism to support the above. The trouble is (for the patent holders), this is in NO WAY how JPEG works. JPEG divides a video stream into blocks (8x8 and 16x16) of pixels, and runs them through a descrete cosine transform. Basically, this turns the representation of the picture into level and percentages of vertical and horizontal waveforms of various frequencies. It then quantizes these values (reducing their size and precision), and orders them from low frequency to high frequency. Then it subjects the whole thing to a run-length algorithm optimized to eliminate zeros (which high quant values tend to do). JPEG is a lossy algorithm that takes advantage of the fact that our eyes don't pick out errors in high frequency components as well as we do low frequency. About the only claim this patent that's similar to JPEG is the Run Length Encoding. But that is covered by prior art that goes back forever. |