Inventing as Solving a Jigsaw Puzzle

Y. B. Karasik,
Thoughts Guiding Systems Corp.,
Ottawa, Canada.

I have to confess that I do not hold patents. Nevertheless I invented a number of things. For example, I invented methods of performing geometric computation with the help of optical devices. I could have taken patents on these inventions but I preferred to publish them as scientific papers.

Now I want to share how I did it in order to enrich and correct TRIZ.

I worked on a software for optimum layout. It turned out that it had to run for days to achieve more or less satisfactory results. I wanted to find a way to speed up the key operations on images. I started with attempts to improve algorithms of their execution on computers. But their run time did not improve much. Then I noticed that our vision performs some of these operations, namely unions and intersections of images, instantly. I started wondering what if I emulate on computer how eye performs these operations. Unfortunately, I did not know if and how eye performs the most time consuming operation called Minkowski sum. Involuntarily I generalized the question as to whether or not it is possible to compute Minkowski sum optically. Having not answered this question I made a further step: what operations on images can be generally performed optically ?

I did not know where to search for answer to this question but recalled that a few years earlier I heard a strange term, "Fourier optics", and even saw a book with such a title. In the absence of any other pointers I reasonably assumed that the book could shed some light on my questions.

In search for the book I headed to the library at my place of work but it did not have it. Instead I found another book, on optical information processing. From this book I learned that besides unions and intersections, which obviously could be performed optically, optics has also means of computing correlation and convolution of images. But how to compute Minkowski sum ? I could not find answer to this question in the book.

I abandoned my effort to develop a software that would emulate the way optics performs the required geometric operations. It did not disappear from my head but was shifted to the background. Instead I resumed developing software for optimum layout based on algorithms of computational geometry.

Years passed during which I encountered other books and articles on optics. One of them, titled "Holographic Methods and Devices for Pattern Recognition", I even bought. But nothing gave me an impetus to resume my effort and point into the right direction.

In March of 1990 I emigrated to Israel and started visiting various research institutions and universities in search for a position. At Tel Aviv University I was introduced to Prof. Micha Sharir who was said to be possibly interested in research on optimum layout. He indeed got interested and offered me the position of ... a Ph.D. student (which was not exactly what I was looking for). But since I did not have other offers, I did not decline it. It was June on the grounds but Sharir promised the position not earlier than October, which lowered my interest in the position even further. We parted having politely exchanged reprints. I gave him a reprint of the only paper on optimum layout that I published in English. He gave me a number of reprints of papers on computational geometry which happened to be in his office.

In a few days I landed the job of a software engineer at a company called "Mennen Medical Ltd". Once in a spare time while working in there I decided to look through the reprints Sharir gave me. One of them, titled "A Kinetic Framework for Computational Geometry" by Leo Guibas et al. immediately attracted my attention because it stated that Minkowski sum is a particular case of convolution. I immediately recalled that convolution can be computed optically, and, hence, through computing convolution one can compute Minkowski sum optically too ! It was a Eureka moment. I finally accidentally found the missing piece of the jigsaw puzzle ! I was so excited that already saw myself having a plant manufacturing and selling optical devices that would perform optimum layout instantly unlike computers which run forever and cannot solve the problem ! To achieve my dream I only needed optical literature to finalize the design. But I lacked it. My only book on optics was left with my parents in Baku. I phoned them and requested to send it me off immediately. By the time it arrived I was already out of work at "Mennen Medical" for better or for worse.

The book had references to such American journals as "Applied Optics" and "Optical Engineering" and I headed to the nearest library (which happened to be in the Weizmann Institute of Science) to read them. Very soon I learned that optical devices to compute Minkowski sum were already built a few years earlier and that I would not make fortunes by employing them to perform optimum layout. "It is difficult to find an untouched stone in science," - my first reaction was. But then I decided to persevere and find a "stone" which was not touched yet. I had a strong feeling that there is one here. OK, I was late with optical setup for Minkowski sum. But what is about convex hull, shortest path, hidden surfaces removal, and other geometric problems ? Examining optical journals revealed that nobody tried to implement a generic geometric optical processor yet. "Then this is what I will do", - I decided.

I spent entire month of October in the library skimming through all issues of "Applied Optics", "Modern Optics", "Optical Engineering" and "Journal of Optical Society of America" in order to compile a list of all operations on images which can be performed optically in one step. By the end of October I had everything to propose the first optical computational model and developed first algorithms in it. Fortunes as a product they did not promise but as research promised to break a new ground. At this point I recalled Sharir's proposition and called him. "Where have you been ?" - he asked me - "the money has been waiting for you for quite some time already."


The invention of a generic optical geometric processor for important practical applications such as image restoration from its boundary, hidden lines removal, etc was of a higher level on the TRIZ scale. Yet there was no contradiction resolved ! For there simply was no contradiction to resolve ! It appears that contradictions are not pertinent to all difficult problems.

Moreover, I did not conduct many trials and errors to arrive at the invention. Of course, I could go to libraries reading everything in a row in search for how Minkowski sum could be performed optically. But I did not do so. Besides TRIZ does not recognize "trying" various books, journals and papers as real trials. By trials TRIZ means trying some ideas of solution.

The whole process of invention in my case resembled solving a jigsaw puzzle. I simply conceived a dream and quickly found all but one blocks to implement it. When the last block had not been found after a certain amount of time, the problem was shifted to the background but my brain remained vigilant to not overlook the missing block should I stumble over it one day. Is it not how many other inventions were made ? Just by accidentally finding a missing block that finalized a jigsaw picture ?

Can it be related to inventions made by accident ? Not, of course. For the idea of the invention was conceived not by accident. Only the final piece of implementation was found by accident. And how any method of inventive thinking could have helped here ?

Even Google would not have helped. Suppose that it existed at that time and that I would have searched for "Minkowski sum" and "Optics". I would have found nothing. But search for "Minkowski sum" alone would have revealed tons of information. To find Guibas et al. paper I had to search for "Minkowski sum" and "Convolution". But how could I know this combination ? One who just associated "Minkowski sum" with "Convolution" would have re-discovered the relationship between them by himself without that paper !

(To be continued)