Seems like old wine in new bottle, HDR using two conversion gain and single readout is not new. OV/ON/Sony already offer such solution with automotive sensors. I think probable innovation here is smaller pixel size and mobile application.
I can't think of an existing HDR implementation that uses dual read out with different conversion gains (as claimed in the video). Typical dual read out HDR schemes use two (or more) integration times.
You all talk in funny code, but any pointers to technical details? I'm not sure how much credence published papers have, I've been told people only publish things that are not actually used in their products, but if someone can, on good authority, point to a paper that comes close to describing this technique, I'd love to look at it. I have seen dual-conversion ideas, but as my brother Anonymous said, these normally involve two separate exposures, and all you avoid is disparate exposure times. If they really managed to get two frames out of one exposure, each with a different conversion, and true signal out of each, that looks like a neat trick.
This blog gets more and more polluted. It truly deviates from its original purpose: inform and relay all image sensor announcements, news, published papers, etc...
Instead, this blog becomes a forum where private, unaware, subjective and useless opinions are expressed.
You could've probably expressed that in a more constructive manner. Vladimir is running a great blog and giving people the forum to ask questions is a wonderful opportunity. However, it is certainly not helpful if the same people asking the questions mistrust the experts' opinions. Brother Anonymous is exaggerating when claiming content of papers isn't used in products. Surely not every trade secret will be disclosed in papers for the competition to copy but that doesn't mean that was is published is nonsense. Samsung doesn't lie in their press announcements and when Vladimir as one of the members of the IISS points to Lofic you should probably be thankful, take this piece of advice and do your own homework...
Some might post anonymous because the companies they work for ask their employees not to publish their work. I have such a nda but still like to ask or comment sometimes here. Great blog, many thanks to Vladimir for filtering the news in such a great way!
A promotional video it is, but pretty thin on the "explains" portion. Does anyone have a relevant technical info on what they are doing here?
ReplyDeleteSeems like old wine in new bottle, HDR using two conversion gain and single readout is not new. OV/ON/Sony already offer such solution with automotive sensors. I think probable innovation here is smaller pixel size and mobile application.
ReplyDeleteI can't think of an existing HDR implementation that uses dual read out with different conversion gains (as claimed in the video). Typical dual read out HDR schemes use two (or more) integration times.
Delete@ I can't think of an existing HDR implementation that uses dual read out with different conversion gains
DeleteWhat about LOFIC?
Ah, yes good point! Maybe the qualifier then is ...for mobile.
DeleteWhat's the smallest pixel pitch reported for LOFIC?
Actually this is not the first time for mobile either: last year one of Xiaomi's flagship handset already used a DCG sensor from another vendor.
DeleteActually it appears similar to the Dual Gain sensor architecture on the Arri Alexa.
ReplyDeleteYou all talk in funny code, but any pointers to technical details? I'm not sure how much credence published papers have, I've been told people only publish things that are not actually used in their products, but if someone can, on good authority, point to a paper that comes close to describing this technique, I'd love to look at it. I have seen dual-conversion ideas, but as my brother Anonymous said, these normally involve two separate exposures, and all you avoid is disparate exposure times. If they really managed to get two frames out of one exposure, each with a different conversion, and true signal out of each, that looks like a neat trick.
ReplyDeleteThis blog gets more and more polluted.
ReplyDeleteIt truly deviates from its original purpose: inform and relay all image sensor announcements, news, published papers, etc...
Instead, this blog becomes a forum where private, unaware, subjective and useless opinions are expressed.
Welcome to Internet 2021!
You could've probably expressed that in a more constructive manner. Vladimir is running a great blog and giving people the forum to ask questions is a wonderful opportunity. However, it is certainly not helpful if the same people asking the questions mistrust the experts' opinions. Brother Anonymous is exaggerating when claiming content of papers isn't used in products. Surely not every trade secret will be disclosed in papers for the competition to copy but that doesn't mean that was is published is nonsense. Samsung doesn't lie in their press announcements and when Vladimir as one of the members of the IISS points to Lofic you should probably be thankful, take this piece of advice and do your own homework...
DeleteIf you think you can do better, I challenge you to start your own blog. I will be the first one to visit your blog.
DeleteSome might post anonymous because the companies they work for ask their employees not to publish their work. I have such a nda but still like to ask or comment sometimes here. Great blog, many thanks to Vladimir for filtering the news in such a great way!
ReplyDeleteLooks like Omnivision OV48C Dual Gain Conversion ...
ReplyDelete