So, the front door of my house was kicked in, and I was robbed. This is not my personal blog, so I will spare you the details; however, I appeal to those who are more knowlegeable in digital manipulation (heh). Here is the shot that was taken of the car that they were driving. I don’t even know where to start, nor do I have the software to try to figure out the plate or any other identifying info. Any help would be appreciated,
I’m extremely well versed in photoshop; I’ll see what I can do, but it doesn’t look like I’m gonna get anything useful at first blush… stay tuned…
Years of watching crime procedurals have taught me that all I need to do is bark “enhance” and a rectangle will appear around the license plate, soon expanding to a magnified inset which will start out blurry but eventually increase in clarity to the point of readability, no matter the resolution of the original photo.
Unfortunately, I don’t have my microphone hooked up to the computer right now, so I’ll have to do it later. 
I’m not an expert at photoshop, but i am fairly well versed in it. It is very hard to blow a picture up and get good resolution. Your best bet might be to up the DPI to 3-400. Save it, then import it back into photoshop and try zooming in again. That might be the best shot you’ve got.
I can’t see squat on the plates except that maybe it starts with an “I”, but I’m pretty sure it’s an Oldsmobile Eighty Eight.
Welp, this is all I could really fudge out of in using Smart Sharpen in Photoshop. Not much help, I don’t think. There just doesn’t seem to be much pixel information there to begin with, and what’s there is pretty soft.
Good luck!
I also tried another method by using the 800 AVE sign as a reference for the degree of sharpening.
It’s odd… I feel like the first digit is a “1” too, but the rest of it seems perhaps purposely obscured? Wonder if they taped over their plate?
I don’t think that the plates are obscured, not sure though. If it helps–a little late I know, but the witness is pretty sure the first number is a 6. Wrong or right, she is the one who took the photo, and given the circumstances she was not looking to be seen. I can’t thank you all enough for looking at this and trying. I wonder if anything in the rear window or the tires/rims would ID this vehicle?
Can you give me more detail on how to do this? I am reasonably intelligent; however I ususally work less with computers and more with troubleshooting things I can touch and feel (read mechanical equipment in manufacturing plants).
I’m sorry, but despite the science-fantasy promulgated by CSI and its ilk, it’s just not possible to what you want to do. There’s no meta-data embedded in your image file; there’s only light on pixels and those pixels are finite. You cannot subdivide the pixels to reveal information that they didn’t capture in the first place. Being obvious here: each pixel is simply a solid square of color; it’s an average of all the colors that were in that dimension/area when the picture was taken. You cannot un-average the colors in the resulting square. All you can do is fudge the resolution between the existing squares. This does not reveal hidden information; that information was never recorded and does not exist within the file.
In the attached blowup, you can see that each pixel contains only one piece of information: the average color of that area. In the blowup within the blowup, you can see the areas were the “missing” information is, and the impossibility of extrapolating that information from what is simply, and only, a single square block of color.
This is what people talk about when they bitch about the bad science in such shows as CSI: it is totally, utterly impossible to increase the amount of information in an image file by blowing it up.
I don’t know what the number-letter convention is for your license plates, but what I can see after some channel manipulation and levels adjustment is “664” or “667”.
Is the lady who sent you the picture computer-savvy enough to have sent you the biggest resolution picture she has? Exporting pictures and uploading them, if you do it automatically, usually makes pics a smaller resolution. But you want the original picture in the original resolution, like it is on her camera or cellphone. Then you can ask Cmyk to work his magic.
Stupid question: You did report this to the police, correct?
I would try this also.
Looking at the camera information embedded in the image reveals that she used a Olympus FE-360 to take the picture, which has a horizontal resolution of 3264 pixels. However, it also shows that she used “digital zoom”, in other words letting the camera crop the resulting image by a factor of 2.8x. This leaves us with 3264/2.8=1167 pixels in the horizontal dimension. The picture you have now is 1024 pixels wide, so she should have a version with slightly more information. That might not be enough to make out the plate, but it’s worth a shot.
I think her most valuable contribution is capturing the dude outside the car. The way he dresses should make him pretty easy to find. :eek:
I don’t think you’ll be getting much out of trying to enhance the resolution – what the camera didn’t capture just isn’t there. However, it’s always worth a try.
I looked whether there might be some other identifying features, and I think that maybe there’s some damage to the rear bumper, either in form of the dangly bit to the left, or, if that’s supposed to be there, in its absence on the right; also, if the white reflection in the car’s left side is from some straight line/object, its crookedness would seem to indicate a dent in that spot. Here’s a picture where I’ve highlighted what I mean. Of course, I might just be seeing things, so take that with a grain of salt.
This may sound stupid but have you shown this to the police?
Their experts could at least id the make and model and coupled with the color could narow it down to a bunch of registered vehicles in the area.
It’s unfortunate that US licence plates aren’t as uniform or plain as those of certain other countries, otherwise it would at least be a theoretical possibility to attempt a solution mathematically, in the same way that one can decipher heavily pixelated representations of MIRC encoding. However given the quality of the image, I doubt this would be possible, or in any even lead to an accurate conclusion.
I pulled this image directly off of her camera. Is there a way for me to pull the other image?
I would like to thank all of your for your help. The burglary has been reported, however the police have to prioritize. I understand that this case is the most important one to me, but they have other more pressing issues. I certainly did not expect for someone to get a full plate clear as day from this; I was hoping to maybe glean some more information to report to the police and my neighbors so we can help protect ourselves. Again I appreciate all of your efforts to help. Thanks!