J-20 5th Gen Fighter Thread VI

Status
Not open for further replies.

siegecrossbow

General
Staff member
Super Moderator
View attachment 51766
Error level around the supposed J-20 is off. In case imgur has recompressed the image, siegecrossbow if you have the image before uploading to imgur, could you perform an ELA on
Please, Log in or Register to view URLs content!
?

I don't think they'd go to the lengths to photoshop a tiny photo of J-20 by the Wuhu airport just to make a point that we've established via satellite image. The poster in question is a reliable PLAAF watcher.
 

Hyperwarp

Captain

SpicySichuan

Senior Member
Registered Member
Please, Log in or Register to view URLs content!


Next stop, Wuhu station.

Gg17I5e.png
Racing with bullet trains? LOL
 

taxiya

Brigadier
Registered Member
View attachment 51766
Error level around the supposed J-20 is off. In case imgur has recompressed the image, siegecrossbow if you have the image before uploading to imgur, could you perform an ELA on
Please, Log in or Register to view URLs content!
?
I have tried FotoForensics couple of times and read its explanation, but failed to understand why it claims that it can distinguish doctored photo from legitimate compressed photos. Maybe my knowledge of JPEG compression is too limited. But from "information theory" that I learnt in school, compression removes information and introduces artifacts, the more compression (high compression rate in one go or multiple compressions) the more artefacts there are, eventually the photo will look unreal with random pixels that make them look doctored. In another word, there is no difference in the end result between doctoring and compression.

The only way FotoForensics' method can make a judgement IMO is by analysing photo in RAW format that involves no compression but only values of pixels from the CMOS/CCD sensors. But most of the photos are JPEG either compressed by the camera (in good quality) or compressed from RAW by Photoshop (still good quality), or repeatedly compressed by various users or photo services (bad quality). I highly doubt tools like FotoForensics can really tell the difference. The centre of the whole thing of "information theory" is that once information is lost in compression, it is lost forever. The artefacts introduced are fake regardless how the end result is reached, it is impossible or pointless to make a difference by the means employed by FotoForensics. Actually, doing heavy or multiple compression is one known practice to make doctored photo looks legitimate, but the opposite of "poor quality photo is likely doctored" can not be determined.

It is possible to tell a doctored photo by examining the angle of light, distortion of geometry of object and wrong perspective etc. But those are not employed by FotoForensics.
 

mys_721tx

Junior Member
Registered Member
The only way FotoForensics' method can make a judgement IMO is by analysing photo in RAW format that involves no compression but only values of pixels from the CMOS/CCD sensors. But most of the photos are JPEG either compressed by the camera (in good quality) or compressed from RAW by Photoshop (still good quality), or repeatedly compressed by various users or photo services (bad quality).

ELA needs compression artifacts to work and does not work on raw bitmaps, or anything that has not been loosely compressed for that matter. JPEG is a lossy compression method and introduces compression artifacts. Each image has its specific error pattern. Digitally combining two images causes inconsistency of the error patterns in the result images. ELA is useful in spotting those inconsistency.

Error level maybe due to image sharpening, like using 'edge sharpening' in PS. I do this most of the time before uploading images.

Agreed. It seems that Adobe products can and will automatically apply sharpening on save. It is quite possible Image Viewer on their phone do this as well. The sharpening may not be intentional on the user's part; however, ELA demonstrate that digital manipulation happened nonetheless.
 

taxiya

Brigadier
Registered Member
ELA needs compression artifacts to work and does not work on raw bitmaps, or anything that has not been loosely compressed for that matter. JPEG is a lossy compression method and introduces compression artifacts. Each image has its specific error pattern. Digitally combining two images causes inconsistency of the error patterns in the result images. ELA is useful in spotting those inconsistency.
Not to continue but just to make clear of my original doubt. I actually got that explanation (in bold texts) from my earlier reading of the website. It is essentially a human (therefor subjective) judgement of what is that specific difference, like what shade of red is yellow, or red or purple, rather than based on some qualifiable number (threshold), so my doubt remains.
 

Hyperwarp

Captain
***

Agreed. It seems that Adobe products can and will automatically apply sharpening on save. It is quite possible Image Viewer on their phone do this as well. The sharpening may not be intentional on the user's part; however, ELA demonstrate that digital manipulation happened nonetheless.

OT: (Apologies to everyone else)
I quite often intentionally modify images. Many people do that. I often change the 'Brightness - Contrast' levels (Hui Tong does that when uploading to his blog) along with 'Color Saturation'. I also use 'Smart/Edge Sharpening' and some other tinkering when images are blurry. As long as you don't doctor the image to make it something that it is not, then it is fine.

For example, sharpening the image (View attachment 51766) so the J-20 is more discernible is ok (but adding non-existing J-20 is PSing). Error level is quite consistent along all edges in that image. IMHO the image is real but the uploader sharpened the edges before uploading. I could be completely off but I think that's what happened.
 

vincent

Grumpy Old Man
Staff member
Moderator - World Affairs
Sadly, I get nothing when i google Bagu and J-16 together.
Nor do I get anything on Global Times website when I search for J-20, AESA, JF-17 or KLJ-7A. :(

bagu = 八股
A slant for official articles that usually provide bits and pieces about some systems
 
Status
Not open for further replies.
Top