Graph image parser

The goal of this project is to produce documented, supportable Unix code (integrating libraries or utilities) which can efficiently and automatically parse quantitative numeric data out of bitmapped graphs in diverse formats, with a minimum of human intervention. We imagine this project breaking down into two phases. Phase One In phase one, our goal is to read the minimum, maximum, and average values from each of a large set (tens of thousands) of bit-mapped images which contain graphs in diverse formats. We first gather any possible values which may already be labeled in the graph image by doing optical character recognition of all bitmapped text found within the image, and searching for labels like “Avg:” or “Maximum:” followed by a numeric value. Next parse the Y-axis of graph, by searching for numeric values stacked vertically along either the left side (usually) or the right side (less commonly) of the graph area, and counting pixels axis-tick-to-axis-tick or text baseline-to-baseline to establish scale. Note that the minimum value may be more or less than zero, and that the value “50,000,000,000” is likely to be abbreviated “50G” or “50 G”. Within the area of the graph, look for sudden RGP value shifts within each vertical column of pixels, and map the “height” of each column of pixels to the quantitative value of that point on the Y-axis. Find the minimum and maximum values from the graph, and then calculate the average value of all readable columns If you have reason to think that it should be more or less expensive than that, I'm happy to hear what you have to say.