DataMatrix参考译码算法.pdf

附件1  Reference Decode Algorithm for Data Matrix This reference decode algorithm finds the symbols in an image and decodes them. This is the decode algorithm used in the determination of symbol quality. This decode algorithm refers to black and white states in the image. The algorithm is actually applied to the normal image and again with black and white states reversed. A. Find the candidate areas which may contain an “L” shaped finder pattern: 1. Select a global threshold midway between the maximum reflectance and minimum reflectance in the image. Convert the image to black and white using the global threshold. 2. Define a distance d which is the length of the shortest expected leg of an imaged symbol’s “L” finder pattern (dwill depend on the pixel density, camera focal length, maximum reading distance, and minimum X-dimension.) 3. Find all the areas with closed continuous boundaries of white on the outside and black on the inside. Note that one candidate area may be totally contained within another candidate area. 4. Disregard all areas which have a longest axis less than 0.5d. 5. Form new merged candidate areas by merging any adjacent areas with a small separation: a. If two areas are separated by less that 0.15d, form a new area by joining the two areas across the minimum separation point. b. Keep the two adjacent areas and the merged area as three candidate areas. c. Repeat the merging process to form new unique merged areas from all original and merged area combinations. 6. Form new divided candidate areas from any candidate areas with a narrow neck separating large sub-areas, ignoring the merged areas from the previous step: a. If an area contains a neck less than 0.15d wide at its narrowest and separates two sub-areas each with a longest axis of at least d, create two new candidate areas