mjpeg once more

alex78 — Fri, 27 May 2005 21:29:02 GMT

Thanks for your previous reply regarding mjpeg.
I am looking forward to your mjpeg sample.

But anyway: I am still confused about the basic principle. I try to decode a continuous flow of jpeg images. Each of these images is a complete jpeg (it includes all jpeg table definitions), but the DQT (Quant. tables) and DHT (Huffman tables) segments of the jpegs are always the same. Your ijl manual says: "Copying the persisted tables to JPEG_PROPERTIES is typically much faster than appending a table stream to the front of each JPEG data stream and forcing the decoder to process and format the tables at every call." So I tried to save and resuse the formatted tables as described in your flashpix sample documentation. But since every jpeg already includes the jpeg tables, how can I tell the decoder not to use them? I also tried to exclude the tables from the jpeg image and to supply the persisted version of the formatted tables, but this does not work for me (I get a one color bitmap as the result). What's wrong? How should I decode this kind of jpeg stream? Thanks

Alex

Re: mjpeg once more

Vladimir_Dudnik — Mon, 30 May 2005 21:51:57 GMT

Hi,

for your information: this technique was developed for work with so called abbreviated JPEG data. It means when JPEG stream does not contain JPEG tables. So you can supply external tables in "raw" or in "formatted" manner.

Regards,

Vladimir

topic Re: mjpeg once more in Intel® Integrated Performance Primitives

mjpeg once more

Re: mjpeg once more