Ismail, I., Hamdy, A., Mostafa, S. (2010). COMPOUND IMAGE SEGMENTATION. JES. Journal of Engineering Sciences, 38(No 2), 533-546. doi: 10.21608/jesaun.2010.124381
Ibrahim Ismail; Alaa Hamdy; Seham Mostafa. "COMPOUND IMAGE SEGMENTATION". JES. Journal of Engineering Sciences, 38, No 2, 2010, 533-546. doi: 10.21608/jesaun.2010.124381
Ismail, I., Hamdy, A., Mostafa, S. (2010). 'COMPOUND IMAGE SEGMENTATION', JES. Journal of Engineering Sciences, 38(No 2), pp. 533-546. doi: 10.21608/jesaun.2010.124381
Ismail, I., Hamdy, A., Mostafa, S. COMPOUND IMAGE SEGMENTATION. JES. Journal of Engineering Sciences, 2010; 38(No 2): 533-546. doi: 10.21608/jesaun.2010.124381
1Professor Doctor Engineer faculty of engineering, Helwan University, Cairo, Egypt.
2Doctor Engineer faculty of engineering, Helwan University, Cairo, Egypt.
3Engineer.
Abstract
Compound document images contain a mixture of natural image and text/graphics. They are very common forms of documents found in magazines, websites, etc. Text and graphics components need special care in the use of compression because text and graphics cannot withstand the significant distortion that is acceptable for natural images. This paper represents a study of different algorithms for segmentation to identify the compound image components. This study focuses on discrete cosine transform (DCT) algorithm, fast Fourier transform (FFT) algorithm, and block based segmentation algorithm; mean, variance, and mean/variance. The segmentation process in general starts with dividing the whole image into non overlapping blocks. Then each block has to be classified either to text/graphics or image via a certain threshold according to the algorithm. the results show that after applying any algorithm the compound image is already classified and the percentage of correct classified text pixels CCPt is calculated and also the percentage of correct classified image pixels CCPi is calculated to present the efficiency of the algorithm, that FFT algorithm is the best one of them that the CCPi between 87.59% in CNN image and 100% in letter image also CCPi between 97.33% TOY image and 100% in leteer image.