Grain counting using neural network method

• After you click ``Count patterns'', the program will pause for several seconds while it searches for candidate patterns in the image or selected region. This may be a long time if the template pattern or the image is large. Color images (including indexed-color images) take considerably longer than grayscale images. It is recommended to test the threshold values on a small part of the image before counting a large image.
• After counting, each occurrence of the pattern that is found is indicated by a red crosshair. If desired, the original image can be restored by clicking ``Restore original''. The program will repeatedly make passes through the image, until no additional matching patterns are found.
• In most cases, the Match Weight should be kept at 1.0. Setting the value lower would cause the area around the pattern (i.e., the mismatched pixels) to have proportionately more influence on whether the pattern is identified.
• Setting the Mismatch Weight to a more negative value may improve the selectivity in some cases, by increasing the penalty for a mismatched pixel. More negative values are needed for complex patterns such as faces.
• When counting grains, the grains must be black on a light background. Use ``Color..Invert colors'' if necessary to achieve this.
• When counting grains, especially on color images, it is recommended to click ``Enhance grains'' before starting. Otherwise, extraneous dark areas (such as cells) may be misidentified as large clumps of grains.
• During the counting, the patterns the most closely match the prototype pattern are recognized first. The counting can be interrupted at any time by clicking on ``Cancel'', pressing Esc, or clicking the main Cancel button.
• When selecting the prototype pattern, select an area of average size and intensity, surrounded by a small region of background. Omitting the background region, or selecting a pattern that is too small, will increase the likelihood that a large dark clump will be identified as several grains that are overlapping.