A New ROI Compression Coding Method Based on SPIHT
|School||Nanjing University of Posts and Telecommunications|
|Course||Signal and Information Processing|
|Keywords||Image Compression Coding Wavelet Transform SPIHT Region of InterestCoding|
In the application of image information, sometimes people pay close attention to some regionsof the image. These areas are the so called ROI (region of interest).Other parts are BG (background). In ROI è we can adopt low loss, or even lossless compression coding to achieve highquality of reconstructed image. But for BG, high compression ratio is executed. In low bit rateconditions, the ROI coding method can ensure high quality reconstruction of important ROI image.Also, the combination of existing image coding algorithms and ROI coding method can solve manyproblems in image processing, such as storage space and bandwidth limitations.There are some defects in the ROI compression algorithms based on SPIHT. The boundarybetween region of interest and back ground is too obvious, and the fusion of the two regions isn’tgood enough. In the SPIHT coding process, the coding methods treat all subbands equally, withoutany coefficient optimization. This may reduce the wavelet coefficients coding efficiency. This thesispresents an improved ROI coding method based on SPIHT. It improves the pre-processing of image,in order that the data integrity of ROI and BG can be maintained, thus the fusion between them maybe more natural. After wavelet decomposition, coefficient optimization is employed in lowfrequency subband, so that the recovery quality of ROI is improved under the same bit rate. In somespecific domain, attention to different areas of the back ground image may be different. Sohierarchical coding is adopted in BG.Through theoretical analysis and experimental simulation, the improved method has beenproved superior to traditional ROI algorithm. In the new coding system, the fusion between ROIand BG image is better. The quality of reconstructed ROI image is higher, and the back groundimage is processed by hierarchical coding, which better meets the demand of practical applications.