Re: Help with jpeg image problem
This is not uncommon for the type of conversion you're doing. If you don't want the banding to occur you'll need to implement a dithering algorithm of some sort.
As far as the single pixels changing from one image to the next (even though it's the same picture) keep in mind that even one bit changes in the 24 bit color can change the last bit in the 16 bit color. Given that each image has thousands of pixels, it's likely that a few, or several, are right at that transition point and even though you say the image is the same they may be slightly different and transition to the other color. I don't have any great solutions for this, unless you can modify the output of the JPEG encoder. An averaging (filter) algorithm might help here, and a general blur with dither will certainly calm the image down at the expense of clarity.
Lastly, double check your setup - many 16bpp LCDs actually use 565 RGB - the green has 6 bits. This may alleviate your banding somewhat if your design has that extra bit.
On 10/2/07, Junior <email@example.com> wrote:
I'm not sure if anyone can offer any advice but it doesn't hurt to try.
I'm using libjpeg to decode a jpeg image stream. My hardware is a 555 16bpp ARM9
(65535 colors) and the image header reads 24bpp RGB and I'm assuming that's 888 format.
When I decode the image, I see the gradient change which I didn't expect even for a
I'm doing everything the docs says and followed the examples with the only difference
being I'm only using the upper 5 bits.
In addition to seeing the well defined color change gradients, I'm seeing a lot
of the image changes pixel values. For example is goes from darkgray to lightgray from one
stream to the other, although the image has not changed.
Has anyone worked with this library and seen this before? Any ideas?
FREE ONLINE PHOTOSHARING - Share your photos online with your friends and family!
Visit http://www.inbox.com/photosharing to find out more!