Image processing spreadsheet

Consider an hypothetical image processing system, where a given target is to be identified. The usual sequence of steps consists on an image equalization to adapt to the illumination situation. Then the image is binarized, allowing a morphological filter to correct the noisy edges and shapes by means of an indeterminate sequence of openings or closings. The resulting image can then be segmented and recognized. If the results are unsatisfactory, then the processing parameters in any of the previous steps must be changed, perhaps by trial and error. For instance, the binarization threshold can be raised or lowered, and the following steps must be performed again to see the results. This is obviously cumbersome, tedious and error prone. The Image Processing Spreadsheet PDICalc is a simple but powerful combination of two different and widespread software technologies. It's benefit comes from enabling users to build an image processing pipeline, considering each step separately, and visualizing the results of modifying the parameters of each step in the final image. A spreadsheet based user interface eliminates the tedious and repetitive interaction that characterizes current image processing software. Users can build a processing template and reliably repeat often needed processing without the effort of redevelopment or recoding. In the cited example the user simply creates the processing template, defining each cell of the spreadsheet as the result of applying a given processing step on another cell. This template can be then reused with any input image, can be stored for future processing sessions, and every step can be trimmed precisely to achieve the desired results. Our implementation considers most of the image processing techniques as its building blocks. Arithmetic operators are overloaded to represent per pixel operations. We included also equalization and histogram correction, arbitrary convolution filtering, arbitrary morphological filtering (with programmed repetition), Fourier operations, and several segmentation techniques.