On 1 May 2018, at 1:45am, Roman Fleysher <roman.fleys...@einstein.yu.edu> wrote:
> If x=10 has less than nX dots, all dots with x=10 are deleted. Because of > deletion, y=3 which previously had more than nY dots no longer passes the > threshold and thus y=3 must be deleted too. This could cause deletion of some > other x, etc. At the end, number of dots on all vertical lines must be more > than nX and number of dots on all horizontal lines must be more than nY. I just realised that this is an image-processing problem, and could be completed ridiculously quickly by defining a filter in a GPU language. If you have access to people working in image recognition, you might usefully consult them about how big a dataset they could support, or just show them the problem in general. I would prototype it using a JavaScript canvas. Once the initial data is written to the canvas you can use API calls to see what colour a pixel is, so the whole thing could be completed without needing further access to the SQLite table. I would bet that any up-to-date browser could solve the problem in JavaScript fast enough for still images. Applying it to successive frames of a moving image might be a different matter. Not helpful for a SQLite solution, I'm afraid. Simon. _______________________________________________ sqlite-users mailing list sqlite-users@mailinglists.sqlite.org http://mailinglists.sqlite.org/cgi-bin/mailman/listinfo/sqlite-users
