Last night I got my ‘blackdensity’ prog debugged to do without error what I have now got it doing. As I have mentioned in previous notes on my website, the prog is finding a suitable value for dParam from a scan presented in greyscale which is presumed to represent a document of text. The basic methodology involves counting fragments for the ‘ideal’ value of the grey-value ‘cutoff’ distinguishing black from white, for a range of trial values for dParam.What I got the prog doing last night was to perform the computation I have just explained briefly, for a number of different resolutions imposed on the greyscale scan. The scanner has supplied an image 1460 pixels in height (by 1024 wide) (and I am using a resolution as regards greyscale of 0 .. 255). To ‘impose’ a resolution corresponding to a document height of (say) 500 simple averaging of greyscale is used through the VB function PaintPicture. Obviously the processing the prog does to find values for dParam and ‘cutoff’ takes longer for higher resolution.
Whether the processing is rendering anything useful (and in nature such processing would, if ‘useful’, decrease the probability of termination of the organism doing the processing and thereby increase the probability of its propagating its genetic composition down the generations) can be estimated by noting whether the fragmentation resulting when the fittest values for dParam and ‘cutoff’ have been found corresponds to analysis of the greyscale field (the scanned ‘document’) into a helpful number of blocks of comparable size. In fact for the resolutions the prog tried (up to a height of 400 cells) nothing helpful resulted. The reason is the scan I was using on this occasion was from a somewhat ‘technical’ book published originally in the 1950s, and in those days publishers (and readers) did not shrink from packing text tightly in a format little different from small-print legal documents.
Thinking further how this type of processing would be useful in nature if a similar situation arose in which the visual information was packed so densely that it could not be processed in any meaningful manner with a single glance, it seems to me this follows: the response of the organism in such a situation would involve shifting the gaze or equivalently ‘zooming’ the attention onto a smaller portion of the visual field in such a way as to reduce the quantity of information in the presentation requiring processing. In the context of reading a document of text a human reader would follow taught conventions, that is in English he would zoom his attention initially at the top of the document and at the left-hand side, and have expectations of an arrangement of text in horizontal lineages.
To briefly bring in the subject of schizophrenia what I would mention is that with schizophrenic brain wiring the reader (or any schizophrenic taker-in of information in any equivalent context) is not led to respond in a successful way to the fed-back demand to reduce the quantity of information up for processing. He is all the time floundering in a sea of excess information from which he can escape only by fleeing (that is, not by restricting his attention there and then in the context he finds himself in). Finding himself in a quiet and/or familiar environment he has less need to flee.
In the context I find myself in (trying to get this computer prog advanced) the way to go is to guide the zooming of attention which the prog must engage in through the principle of locating blocks of text in a scan which may represent a page in a newspaper or magazine. One little practical point I may mention here is that in locating blocks of text a human reader does not only consider the outline (that is, looking for horizontal lineages) but also samples from small subregions of the visual field thrown up from automatic pre-processing done by packets of neurons within the retina, and in this case helpful if they report the presence of blocks looking like letters of the alphabet.
