엔트로피 코딩을 위한 적응적 예측기

Abstract

Listen

Translate article

Abstract In this paper, an efficient predictor for entropy coding is proposed. It adaptively selects one of two prediction errors obtained by MED(median edge detector) or GAP(gradient adaptive prediction). The reduced error is encoded by existing entropy coding method. Experimental results show that the proposed algorithm can compress higher than existing predictive methods.요 약 본 논문에서 엔트로피 코딩을 위한 효과적인 예측기를 제안한다 . 제안하는 예측기는 MED(median edge detector)와 GAP(gradient adaptive prediction)의 예측 에러 중의 하나를 적응적으로 선택한다. 감소한 에러는 기존의 엔트로피 코딩 방법을 이용하여 부호화한다. 실험 결과, 제안하는 알고리즘이 기존 예측 방법보다 향상된 압축이 가능함을 보인다. Key Words : Predictor, Entropy Coding, Lossless, Image Coding * Corresponding Author : Hyun-Sang Park(vandammm@kongju.ac.kr)Received October 1, 2009 Revised (1st December 11, 2009, 2nd December 20, 2009) Accepted January 20, 2010 1. Introduction Recently, many researches for image compression of digital images are increased. Especially, lossless compression is an important field of application for image compression. High-end digital devices enable the user to access the raw, uncompressed picture, i.e. not altered by any coding algorithm. Many algorithms were proposed. Context-based adaptive prediction schemes [1-5,7,8] have shown significant improvements over fixed prediction schemes. CALIC [1] uses gradient adaptive prediction (GAP). The new lossless compression standard JPEG-LS [2] adopts median edge detector (MED). A simple data prediction technique such as DPCM can de-correlate image data in smooth areas with very low computational cost. Prediction can be viewed as a context modeling technique of very low model cost that is highly effective under an assumption of smoothness. In JPEG-LS and CALIC, they chose to employ MED predictive coding and GAP predictive coding, respectively.In this paper, we propose an efficient technique called adaptive prediction algorithm which selects one of results obtained by the MED and GAP prediction, properly. Thus, it can reduce prediction error and obtain the reduced entropy of the residual error. Our proposed prediction method achieves good performance for entropy image coding and outperforms existing methods, such as MED and GAP, while having a low complexity.