Restoration of Degraded Document Images ppt

Embed Size (px)

Citation preview

  • 7/21/2019 Restoration of Degraded Document Images ppt

    1/24

  • 7/21/2019 Restoration of Degraded Document Images ppt

    2/24

    Computer originated documents are main formtransmitting information, and Document ImageProcessing (DIP) deals with the conversion of textdocument images into computer readable format.

    Two phases of DIP: Binarization and CharacterRecognition.

    Binarization is the key process, as characterrecognition significantly depends upon it.

  • 7/21/2019 Restoration of Degraded Document Images ppt

    3/24

    It is also known as thresholding.

    It is the process of separating for foreground textregion from the background surface.

    For uniform background, binarization is easy due tothe fact that there is a distinct separation betweenforeground and background.

    For non-uniform background or noisy background,some special binarization techniques are required.

  • 7/21/2019 Restoration of Degraded Document Images ppt

    4/24

    Binarization are broadly classified into two categories-global and adaptive.

    In global binarization, single threshold value is used

    for separation of foreground from background. It gives poor performance under noisy environment.

    In adaptive binarization, threshold value is selectedadaptively, hence, gives better performance for

    binarization of degraded document images. Adaptive binarization is used in degraded DIP.

  • 7/21/2019 Restoration of Degraded Document Images ppt

    5/24

    The entire document image is scanned by a smallwindow. Window size may vary for different algorithm

    Threshold value is selected based on the pixelinformation under the window. Pixel information

    generally include some statistical information such asmean, standard deviation etc.

    The statistical information are different for differentbinarization techniques.

    Thus, we get different threshold for different windowposition.

  • 7/21/2019 Restoration of Degraded Document Images ppt

    6/24

    As noise is generally present in the background surfaceof document images, adaptive binarization can handlethe noise.

    The performance of the binarization depends on thethreshold value of the degraded portion of thedocument.

  • 7/21/2019 Restoration of Degraded Document Images ppt

    7/24

    The resulting binary image is processed by this phase.

    The binary image is first segmented, and thencharacter recognition is performed.

    This character recognition can be done by OpticalCharacter Recognition (OCR) process.

    The performance of the resulting binary image ofdifferent binarization techniques can be evaluated inOCR.

  • 7/21/2019 Restoration of Degraded Document Images ppt

    8/24

    Four different adaptive binarization methods areevaluated- Niblacksmethod, Sauvolasmethod, Gatos

    et al.smethod and finally Halabi et al.smethod.

    The performance of each method is evaluated by OCR.

    A comparison is made between the OCR results of

    different techniques.

  • 7/21/2019 Restoration of Degraded Document Images ppt

    9/24

    Niblacks method is based on the pixel informationbased on the local mean and standard deviation of thepixel under the window.

    The threshold value is calculated by the followingformula:

    T = m + k * s

    where, m is the local mean and s is the standarddeviation; k is constant usually set to -0.2.

  • 7/21/2019 Restoration of Degraded Document Images ppt

    10/24

    This is slightly modified version of Niblacksmethod.

    It considers the dynamics of standard deviation along

    with the mean and standard deviation. The threshold value is given by the following formula:

    T=m * ( 1 k * ( 1 s/R))

    where, R is the dynamics of standard deviation. Forgrayscale images, the value of R is set to 128.

  • 7/21/2019 Restoration of Degraded Document Images ppt

    11/24

    This method uses Sauvolasbinarization technique asan intermediate step.

    It consists of five distinct steps:

    1. Preprocessing2. Rough estimation of foreground text region

    3. Approximate background surface calculation

    4. Final thresholding

    5. Post processing

  • 7/21/2019 Restoration of Degraded Document Images ppt

    12/24

  • 7/21/2019 Restoration of Degraded Document Images ppt

    13/24

    In DIP, a preprocessing step of documents are usuallyperformed for reducing the degree of noise.

    Generally filtering of document images reduces noise.

    Gaussian filter, average filter, median filter can be usedfor this purpose.

    In this method, an adaptive Wiener filter is used.

    The window size of the filter depends upon the several

    factors such as size, thickness of the text character. This process causes blurring effect of the original

    image.

  • 7/21/2019 Restoration of Degraded Document Images ppt

    14/24

    Sauvolasmethod is used for this step.

    The pixel value corresponding to the text region givesthe superset of the correct text pixels.

    The wiener filtered grayscale image is binarized bysauvolasmethod, all 0s in the output imagecorresponds to the text regions.

  • 7/21/2019 Restoration of Degraded Document Images ppt

    15/24

    A background surface is produced using the filteredgray scale image and the roughly estimated text region.

    The pixel value in the resulting image correspondingto the non-text region in the roughly estimatedforeground image remains same as the original wienerfiltered gray scale image.

    The pixel value in the resulting image corresponding

    to the text region in the roughly estimated foregroundimage is determined by interpolation of theneighborhood pixels around it.

  • 7/21/2019 Restoration of Degraded Document Images ppt

    16/24

    Final thresholding is done by combining the estimatedbackground surface with the filtered grayscale image.

    A threshold value is determined by calculating the

    distance between pixel of the original image and theestimated background image.

    The pixel value corresponds to text region if thedistance is greater than the threshold.

    The threshold changes according to the pixel value ofthe grayscale background image.

    It preserves the textual information in presence ofdegraded background.

  • 7/21/2019 Restoration of Degraded Document Images ppt

    17/24

    Post-processing step is required for the furtherreduction of noise level at the output binary image.

    Post processing step is done using shrink and swellfilter.

    Shrink filter examines each foreground pixel, if thenumber background pixel under the window is greaterthan some threshold value, the foreground pixel isconverted to the background pixel.

    This results in thinning of the character as well as itreduces noise (salt and pepper) in the backgroundsurface.

  • 7/21/2019 Restoration of Degraded Document Images ppt

    18/24

    Swell filter examines every background pixels, if thenumber of foreground pixel is greater than thethreshold, the background pixel is converted to the

    foreground pixel. It fills the holes in the text as well as thickness of the

    text character is increased.

    The size of the filtering window and the thresholdvalue depends on the average size of the character

  • 7/21/2019 Restoration of Degraded Document Images ppt

    19/24

    It is almost same as the previous method except it usesGaussian filter in the preprocessing step.

    Due to the use of the gaussian filter, blurring of edgesoccur.

  • 7/21/2019 Restoration of Degraded Document Images ppt

    20/24

  • 7/21/2019 Restoration of Degraded Document Images ppt

    21/24

    At Niblacksmethod, the binarization performs well inthe areas near to the text region. However, largeamount of noise is observed in non-textual portions.

    This is the main drawback of Niblacksapproach. At Sauvolasmethod, the binarization performance is

    good in presence of noise such as non uniformbackground, low contrast etc. It solves the problem of

    Niblacks method i.e. the noisy binarization of non-textual region by taking care of the dynamic range ofthe grayscale background. But as the degradationbecomes severe in background, the resultingbinarization performance decreases.

  • 7/21/2019 Restoration of Degraded Document Images ppt

    22/24

    The method proposed by Gatos et al. performs well inpresence of degraded background. The thresholdcalculation depends on the estimated backgroundsurface and binarizes the text image by taking care of

    the degradation in the background. The method proposed by Y.S. Halabi et al. also

    performs well in presence of degradation. TheGaussian function used in the preprocessing step takescare of the noise present in the original grayscale textdocument image. However, blur of edges occur due theGaussian function and information loss may occurbecause of the blurring effect.

  • 7/21/2019 Restoration of Degraded Document Images ppt

    23/24

    Gatos B., I. Pratikakis, S. J. Perantonis, Adaptive DegradedImage Binarization, Journal of Pattern Recocnition, 39(2006)

    Sauvola J., M. Pietikainen, Adaptive Document ImageBinarization, Pattern Recognition 33 (2000)

    Basilios Gatos, Ioannis Pratikakis and Stavros J. PerantonisAn Adaptive Binarization Technique for Low QualityHistorical Documents (2004)

    Yahia S. Halabi, Zaid SA, Faris Hamdan, Khaled Haj YousefModeling Adaptive Degraded Document Image

    Binarization and Optical Character System (2009) W. Niblack, An Introduction to Digital Image Processing,

    Prentice-Hall, Englewood Cliffs, NJ, 1986 ABBYY (www.finereader.com).

    http://www.finereader.com/http://www.finereader.com/
  • 7/21/2019 Restoration of Degraded Document Images ppt

    24/24