its mri brain image and it goes to any size and any shape
but I boarder it with chav-nese segmention and u see a result
What you are asking to do cannot be done under the conditions you describe.
"Each of the pictures has something in it that can be any size or shape or color, but I know it to see it. The object has to be automatically detected."
Raymond Smullyan wrote in his book "This Book Needs No Title",
<<The story is of a Zen-master who was delivering a sermon to some Zen monks outside his hut. Suddenly he went inside, locked the door, set the hut on fire, and called out, "Unless someone says the right thing, I'm not coming out." Everybody then desperately tried to say the right thing and, of course, failed. Along came a latecomer who wanted to know what all the fuss was about. One of the monks excitedly explained, "The master has locked himself inside and set fire to the hut and unless somebody says the right thing, he won't come out!" Upon which the latecomer said: "Oh my God!" At this the master came out. [...]
Obviously, what the master wanted was a response which was wholly spontaneous. Any response which was "designed" to persuade the master to come out would have ipso facto failed. When the latecomer said, "Oh my God!" he was not "trying to say the right thing"; he did not say it with any idea that it would bring the master out; he was simply alarmed! Had anyone else present said "Oh my God!" for the purpose of getting the master to come out, the master would probably have sensed it and accordingly would not have come out.>>
The point of the above in context is that there can be things of no fixed size or shape or color that you, in your thoughts, can recognize and point out, and yet which cannot be found by any automated system. If there is something that you want to be recognized by an automated system, you have to put boundaries on what it can look like.