Methods and apparatus for auto image binarization
A threshold determination method is selected from among a plurality of alternative global thresholding determination methods and, optionally, a local thresholding determination method based on characteristics of a histogram of grayscales values representing an image. When it is determined to use a global thresholding method, a single global binarization threshold value is determined using the selected global thresholding method. Various alternative global binarization threshold values include a predetermined constant, an average value of the two grayscale values, an Otsu method based threshold value, a Newton method based threshold value, and an Otsu method based threshold value based on a truncated version of the histogram. When it is determined to use local thresholding, a plurality of local binarization threshold values are determined corresponding to different non-overlapping blocks of the image. The determined binarization threshold(s) are applied to the gray scale pixel values to obtain a set of binary pixel values.
1. An image binarization method, the method comprising:
generating statistical information indicating a distribution of pixel values for a set of pixel values representing an image;
selecting between a plurality of global binarization threshold determination methods and a local binarization threshold determination method as a function of said generated statistical information; and
generating a set of binary pixel values from said set of pixel values representing said image.
2. The method of claim 1 , wherein said generating statistic information includes generating a histogram from said set of pixel values.
3. The method of claim 2 , wherein said plurality of global binarization threshold determination methods includes a first method which uses a truncated version of said generated histogram and wherein a second one of said plurality of global binarization threshold determination methods is a Newton based threshold determination method.
4. The method of claim 1 , wherein selecting between a plurality of global binarization threshold determination methods and a local binarization threshold determination method includes:
determining a number of spikes in said histogram; and
selecting a threshold determination method as a function of the determined number of spikes in said histogram.
5. The method of claim 4 , wherein selecting a threshold determination method includes selecting a first global threshold determination method when said number of spikes equals one, said first global threshold determination method including using a global binarization threshold value.
6. The method of claim 5 , wherein said global binarization threshold value is a middle value in a range of possible pixel values.
7. The method of claim 5 , wherein selecting a threshold determination method includes selecting a second global threshold determination method when said number of spikes equals two, said second global threshold determination method including using a second global binarization threshold value determined based on the pixel values corresponding to said two spikes in said histogram.
8. The method of claim 7 , wherein selecting a threshold determination method includes comparing a histogram entry corresponding to the second largest spike to a histogram entry threshold; and
when the histogram entry corresponding to the second largest spike is larger than the histogram entry threshold using a third global binarization threshold value determined based on the pixel values corresponding to the two largest spikes in said histogram.
9. The method of claim 8 , wherein said third global binarization threshold value is equal to said second global binarization threshold value.
10. The method of claim 8 , wherein when the histogram entry corresponding to the second largest spike is less than or equal to the histogram entry threshold said global binarization threshold value is used as a global binarization threshold.
11. The method of claim 7 , wherein selecting a threshold determination method includes selecting a global threshold determination method when said determined number of spikes is greater than two and less than a first threshold value.
12. The method of claim 11 , wherein said first threshold value is 12.
13. The method of claim 11 , further comprising:
when said determined number of spikes is greater than or equal to said first threshold value,
processing said histogram to generate a normalized histogram;
calculating a kurtosis from the normalized histogram;
comparing the kurtosis to a kurtosis threshold; and
determining to use a global binary threshold determination method when the comparison of the kurtosis to the kurtosis threshold indicates that the kurtosis exceeds the kurtosis threshold.
14. The method of claim 13 , wherein when the kurtosis exceeds the kurtosis threshold the determined method of generating a global binary threshold includes:
calculating a global threshold, said global threshold being one of Newton method based threshold and Otsu method based threshold.
15. The method of claim 13 , wherein when the kurtosis does not exceed the kurtosis threshold, the method further comprising:
determining a number of clusters in the histogram; and
when the number of clusters exceeds a cluster threshold,
determine to use a local binary threshold determination method; and
determine local threshold values to use for binarization of different blocks of said image.
16. The method of claim 15 , wherein the local threshold values are determined using the Otsu method.
17. An apparatus for image processing comprising:
a processor;
a module for generating statistical information indicating a distribution of pixel values for a set of pixel values representing an image;
a module for selecting between a plurality of global binarization threshold determination methods and a local binarization threshold determination method as a function of said generated statistical information; and
a module for generating a set of binary pixel values from said set of pixel values representing said image.
18. The apparatus of claim 17 , wherein said module for generating statistical information includes a module for generating a histogram from said set of pixel values.
19. The apparatus of claim 18 , wherein said plurality of global binarization threshold determination methods includes a first method which uses a truncated version of said generated histogram and wherein a second one of said plurality of global binarization threshold determination methods is a Newton based threshold determination method.
20. A non-transitory physical computer readable having code stored thereon, the non-transitory physical computer readable medium comprising:
code for causing a computer to generate statistical information indicating a distribution of pixel values for a set of pixel values representing an image;
code for causing the computer to select between a plurality of global binarization threshold determination methods and a local binarization threshold determination method as a function of said generated statistical information; and
code for causing the computer to generate a set of binary pixel values from said set of pixel values representing said image.