Scalable reduced dimension object segmentation based adaptive progressive color-image coding

Abstract

A fast and efficient wavelet image-coding algorithm based on scalable reduced dimension object segmentation (SRDOS) introduced by L. Zhang et al. (2003) is presented in this paper. The object is segmented with the lifting scheme right after wavelet transform. It is the reduced dimension space of the transform domain in which SRDOS algorithm is applied to detect and segment the object with lower complexity and more sufficient accuracy. Due to the characteristics of multi-scale segmentation and higher performance/complexity, SRDOS adapts to object-oriented adaptive progressive wavelet color-image coding. For higher compression ratio, coding algorithm takes advantage of the relationship of the same spatial location at the same scale to remove spatial redundancy of the landscape orientation and applies context-based arithmetic coding algorithm to reduce spatial redundancy of the vertical orientation. For keeping important information, object and non-object have different quantizers and priorities respectively. The experiments show that SRDOS based coding algorithm gains higher subject visual quality and image peak signal to noise ratio (ISNR).