This is to certify that the project entitled STUDY OF APPROACHES FOR REMOVAL OF ADVISORY AND OVERLAPING IN DOCUMENT IMAGES submitted by Pritish Sahu No: 107CS044 partially fulfills the requirements for the degree of Bachelor of Technologist in Computer Science and Engineering at the National Institute of Technology Rourkela authentic work done by him under my supervision and guidance. To the best of my knowledge, the subject matter included in the thesis has not been submitted to any other University/Institute for the award of any degree or degree. Although it would be easy to name them all, it would not be easy to thank them enough.
Pankaj Kumar Sa, Department of Computer Science Engineering, for their valuable guidance, constant encouragement and kind help at different stages for carrying out this thesis work. We would like to thank all our friends who helped us and also want to thank everyone who directly or indirectly contributed to the success of our work. This project implements algorithms to reduce throughput distortion using techniques in digital image processing.
We study the algorithm knowing that in images the high frequency components are sparse and stronger on one side of the paper than on the other. Bleed-through effect and show-through effect were removed at once without repetition. Here, the sources need not be required to be independent or the mixture to be invariant. Therefore, it is suitable for separating mixtures such as those produced by blow-through.
Image processing is any form of signal processing for which the input is an image, such as a photograph or a video frame; the output of image processing can be either an image or a set of parameters related to the image.
Image Processing
For the first seven categories the inputs and outputs are images, while for the other three the outputs are attributes of the input images. With the exception of image acquisition and display, most image processing functions are implemented in software. In image processing, the technique that works well in one field may be unsuitable in another, as it is characterized by specific solutions. Of the above ten subcategories of digital image processing, this thesis deals with wireless and multi-resolution processing as well as image enhancement.
Spatial domain methods, this refers to image plane itself and approaches in this category are based on direct manipulation in an image. Frequency domain methods, which operate on the different types of transformation (fourier transform, short fourier transform, wavelet transform) of an image.
Document
After applying these text and graphics analysis techniques, the several megabytes of initial data are stripped to provide a much more concise semantic description of the document.
Problem Definition
Motivation
Thesis Organization
Conclusion
Most of the signals we use are TIME-DOMAIN signals in their basic format. When we plot time-domain signals, we get a time-dependent (usually amplitude) representation of the signal. In many cases, most distinguishable information is hidden in the frequency content of the signal. The Fourier transform tells us a lot about a signal, it tells us what frequency components exist in the signal. We measure frequency using the Fourier transform.
Fourier Transform
Definition
Working Mechanism
Why fourier transform is not suitable?
Short Term Fourier Transfrom
Definition
Mathematical Approach
Resolution Issues
Conclusion
Wavelet transforms are based on small waves called wavelets with varying frequency and limited duration. It also helps with MRA (multi-resolution analysis), i.e. note that the sinusoid has an easily observable frequency, while the wavelet has a pseudo-frequency in the sense that the frequency varies slightly over the length of the wave[5].
Why Wavelet?
Continuous Wavelet Transform
Stationary Wavelet Transform
What swt does?
Filtering a signal corresponds to the mathematical operation of convolution of the signal with the impulse response of the filter. A low-pass half-band filter removes all frequencies that are above half the highest frequency in the signal. The signal is then down-sampled by 2 since half the number of samples are redundant.
At each level, filtering and subsampling will result in half the number of samples (and thus half the temporal resolution) and half the frequency bandwidth (and thus double the frequency resolution) [4].
Conclusion
The stationary wavelet transform (SWT) is a wavelet transform algorithm designed to overcome the lack of conduction invariance of the discrete wavelet transform (DWT). The stationary wavelet transform (SWT) is mainly used for denoising. Here, we studied methods for discrete non-linear image mixtures. The study involves digitizing both sides of the document and attempting to completely align both sides by hand.
Initial Process
Separation Methods
Algorithm
PCA
Compute the difference between the empirical mean of the vector u from each column of the data matrix X and store the data in the matrix M N B. The sum of the energy content of all eigenvalues from 1 to m gives the cumulative energy content g for the m-th eigenvector . Take the square root of each element along the main diagonal of the covariance matrix.
After calculating PCA we need to calculate the value to which level it will be decomposed using SWT. The main motive behind the procedure is that it preserves the coefficients that are stronger in the mixture component than in the opposite component, and weakens the coefficients that are weaker than in the opposite component. The calculation process described in was applied to all horizontal, vertical and diagonal wavelet coefficients at all decomposition levels where Hj-horizontal coefficients at level j, Vj-vertical coefficients at level j,Dj-diagonal coefficients at level j.
The reconstruction of the separated images was performed using the wavelet method using the high-frequency coefficients (Hj, Vj, Dj) after competition. For the low-frequency coefficients (Δn in Figure 9) The coefficients obtained from the decomposition of the corresponding mixture image were used without change. The work in this thesis focuses primarily on removing bleeds and bleed-through in scanned images.
Using SWT(Stationary Wavelet Transform) and Haar wavelet, we removed bleed-through and bleed-through from scanned images. It is a faster method with no iteration needed to separate real nonlinear mixture of images. Here we use wavelet decomposition method to a deep level, the results are strong non-point wise separation. Unlike previous solutions, this method does not assume that the mixture is invariant, and is therefore suitable for mixtures with different local properties, such as those resulting from bleed-through or from wrinkled documents. The following limitations were encountered during the course of this project: The output obtained from this method applied to color images is not good, better results could be obtained by processing the three color channels of the original images separately.
The calculation of contrast compensation becomes complex and can be evaluated in future work. Color images can also be captured by separately processing the three color channels of the original images.
