I have tried also with text and a lot of dots, but I found out that the resolution must be super hight for the effective work of tesseract. I`ll be glad if someone can help. If someone think of another way - please to share his opinion, but in that case there is one more condition to observe - on the page there can be not only dots, but text and raster images, so the algorithm should make difference between these three objects, and get only the dots.