Designing a CODEC System for High-Resolution Textual Images Based on Super-Resolution
Paper ID : 1136-MVIP2020 (R1)
Saeid Moradi *1, Hadi Grailu2
1Electric & Robotic Engineering Department, Shahrood University of Technology
2Electrical & Robotic Engineering Department, Shahrood University of Technology, Shahrood, Iran.
In this paper, a CODEC system for high-resolution textual images based on super resolution has been proposed. The proposed method has used the idea of dynamic range reduction in textual images in order to achieve more compression ratio. Dynamic range reduction along with the compression may reduce image quality. Therefore, a method should be chosen in the reconstruction unit that can both increase the image size and modify undesirable effects that affect the image. In this paper, in the reconstruction phase, interpolation-based super-resolution method has been used. In this method, first, a highly efficient textual image matting algorithm based on local linear modeling has been used to decompose the low-resolution input image into three layers. Then each layer is up-sampled with a particular method based on its information importance. Finally, these three high-resolution layers have been composed to obtain the high-resolution image. An important feature of the proposed method is the ability of combining it with various compression methods. The combination of the proposed method with the JPEG, the JPEG2000 and the SPIHT compression methods has been considered. The proposed method could increase compression ratio for the JPEG and the JPEG2000 compression methods. Although an acceptable answer has been obtained in terms of optical character recognition (OCR) and mean opinion score (MOS), in terms of peak signal to noise ratio (PSNR) the proposed method has not had better result than other techniques.
textual image compression, JPEG compression, JPEG2000 compression, SPIHT compression, super-resolution, optical character recognition.
Status : Paper Accepted (Oral Presentation)