How to compare two pictures?

E

Evgeny2017-01-12 22:06:00

PHP

Evgeny, 2017-01-12 22:06:00

There is a paper document that fills out over time, it needs to be periodically scanned and sent to us.
It is necessary to find the previous scan in the database and replace it with a new one.
The format is simple - table, first column - full name. by hand, the first line is the date and time. everything else is marks
Thoughts are as follows: we form a picture according to this picture, into which only the part that does not change falls - i.e. the first 10 names, the title (it is written by hand on paper) and then we look for a copy of this picture in the table of the same pictures for other scans.
When googling found a perceptual hash. However, I am not sure that he will be able to cope with the task well enough.
Actually tell me in which direction to dig to find the previous version of the scan?
m.b. who faced a similar problem?

Reply

Answer the question

In order to leave comments, you need to log in

3 answer(s)

T

ThunderCat, 2017-01-12
@ThunderCat

It's just magical, that is, there are people who think that if they scan 2 times 1 sheet, they will get 2 matching images ... Hmm, interesting ...
The images will be very different, you scan the sheet, then they write / dance / drink on it coffee, shoved into the scanner again, but from a different angle, scanned and compared ... Yes, just scanned - pulled out - put in will be seriously different. A function that will RECOGNIZE all this and at the same time shovel all the pictures that are there - this is clearly some kind of hellish hell.
Isn't it easier for each document to have a record in the database/file with id, document name, file name, date and comments, and store all this normally???

V

Vitaly, 2017-01-12
@vshvydky

look at https://www.npmjs.com/package/jimp
there is a diff method, you can sprinkle images, but in general a piece of paper with checkmarks by hand, it's a gesture ....

D

dom1n1k, 2017-01-12
@dom1n1k

A perceptual hash is definitely not good - it is more suitable for photographs and the like than for forms filled with squiggles.
But generally speaking, the problem statement is incomprehensible. Need to compare just on the basis of equal / not equal? Or understand what exactly has changed? And why can't you search for scans stupidly by file name and date (which can also be included in the name)?