|
|
Post by rob260259 on May 1, 2011 16:23:38 GMT -4
|
|
|
|
Post by Mr Gorsky on May 2, 2011 4:53:44 GMT -4
I have no idea what "deconvolved" means, but nice video ... thanks for the link.
;D
|
|
|
|
Post by chrlz on May 2, 2011 6:40:06 GMT -4
Allow me to be shot by purists for this 'explanation'..
*Convolution* is the sum of the blurring and other distortions that are captured in any image. DEconvolution is the name for any process that tries to reduce or counteract that - in this case that would be a computer program which understands what caused the blurring/distortions, and uses algorithms to 'undo' them and provide a slightly clearer/sharper image.
Emphasis on *slightly*. It has to be very carefully done, has lots of limitations, and usually only gives small improvements.
|
|
|
|
Post by echnaton on May 2, 2011 11:20:02 GMT -4
That sounds a lot like using Auto-Tune to correct poor pitch for a singer. Just a few touches can save you hours of studio time and tedious rerecording of vocals, too heavy of a use and you sound like Cher. Allow me to be shot by purists for this 'explanation'.. *Convolution* is the sum of the blurring and other distortions that are captured in any image. DEconvolution is the name for any process that tries to reduce or counteract that - in this case that would be a computer program which understands what caused the blurring/distortions, and uses algorithms to 'undo' them and provide a slightly clearer/sharper image. Emphasis on *slightly*. It has to be very carefully done, has lots of limitations, and usually only gives small improvements. |
|
|
|
Post by lukepemberton on May 2, 2011 11:37:00 GMT -4
That sounds a lot like using Auto-Tune to correct poor pitch for a singer. Just a few touches can save you hours of studio time and tedious rerecording of vocals, too heavy of a use and you sound like Cher. You mean a bit like this. GoneToPlaid has done a great deal of work in this area, not only with the LRO images, but general debunking of 'fake photo' claims and pointing out Prof Colin Rourke's errors. Many of his YouTube videos are worth looking at, along with his website. www.youtube.com/user/GoneToPlaidwww.apollo.mem-tek.com/ |
|
|
|
Post by BertL on May 2, 2011 11:56:43 GMT -4
That sounds a lot like using Auto-Tune to correct poor pitch for a singer. Just a few touches can save you hours of studio time and tedious rerecording of vocals, too heavy of a use and you sound like Cher. I think it would be more suitable to compare it to tools that filter out white noise or background noise, or get rid of those high pitched squeaks you hear sometimes with live microphones. |
|
|
|
Post by ka9q on Jun 14, 2011 22:07:48 GMT -4
Allow me to be shot by purists for this 'explanation'.. *Convolution* is the sum of the blurring and other distortions that are captured in any image. DEconvolution is the name for any process that tries to reduce or counteract that - in this case that would be a computer program which understands what caused the blurring/distortions, and uses algorithms to 'undo' them and provide a slightly clearer/sharper image. Another analogy would be to the use of an audio equalizer. Let's say you have a musical recording made with microphones that have a very uneven frequency response. The recording is that of the original sound convolved with the "impulse response" of the microphone. To correct this recording with an equalizer, you'd program in the inverse of the microphone's frequency response; where the microphone was 5 dB low, you'd set the equalizer to +5 dB, and so on. The result would be a recording with a flat frequency response. This method has its limits; it can't recreate information that simply isn't there. If the microphone had no response at a given frequency, then there's nothing on the recording to work with; if you cranked the equalizer at that frequency way up you'd only boost whatever noise is there. Image and audio processing share many of the same signal processing tools. The main difference is that while audio signals are generally 1-dimensional, images are 2D. Fortunately, there are 2-dimensional extensions of most 1D algorithms. |
|
