用於混合式向量量化之編碼器設計

Abstract

近年來基於向量量化用於數位影像編碼的技巧, 已提供一種能用於低位元 比率影像編碼之途徑, 並且保持良好影像品質. 但所負出之代價是複雜的 運算. 而實際上也有許多方法來減輕此種負擔. 其中一種結合離散餘弦轉 換與向量量化(DCT/VQ)及區塊截斷編碼與向量量化(BTC/VQ)的方法被提 出. 此種方法能夠產生高畫質高壓縮率的結果, 而且能使向量量化之向量 簿的向量數目減少. 也因如此, 所以能夠降低在編碼時之運算複雜度. 我 們將此方法稍為修正, 使其能適合於VLSI之設計, 但是仍可保持高畫質及 高壓縮比. 在此提出之架構中, 我們不使用乘法器.其中內部所需的離散 餘弦轉換轉而以一個部份離散餘弦轉換(PDCT)來趨近.若提供50百萬赫茲 的時脈, 而影像為每秒30張之512x512大小, 則此架構可運用於即時影像 編碼. Digital image coding using vector quantization(VQ) based technique provides low-bit rates and high quality coded images , at the expence of computational demands. Many approaches have been used to alleviate the encoding search process. A novel method with DCT/VQ and BTC/VQ combined can chieve a high quality and low-bit rate compression of images. Since this method has the effect of reducing the codebook size, thus it has alleviated the encoding search loading. We modified this method for VLSI implementation. It preserves the good results in image quality and compression ratio and provides efficient architecture design. We designed an encoder for this modified method. In this design we do not use any multiplier element. The DCT part is approximated by some adders and shifts, and performs the partial discrete cosine transform(PDCT). With 50MHz clock being provided , this circuit can provide real-time encoding of 512x512 frames at 30 frames/sec frame rate.